Genome evolution in Arabidopsis/Brassica: conservation and divergence of ancient rearranged segments and their breakpoints


  • Piotr A. Ziolkowski,

    1. Department of Biotechnology, Adam Mickiewicz University, Międzychodzka 5, 60-371 Poznań, Poland, and
    2. Institute of Plant Genetics, Polish Academy of Sciences, Strzeszyńska 34, 60-479 Poznań, Poland
    Search for more papers by this author
  • Malgorzata Kaczmarek,

    1. Institute of Plant Genetics, Polish Academy of Sciences, Strzeszyńska 34, 60-479 Poznań, Poland
    Search for more papers by this author
  • Danuta Babula,

    1. Institute of Plant Genetics, Polish Academy of Sciences, Strzeszyńska 34, 60-479 Poznań, Poland
    Search for more papers by this author
  • Jan Sadowski

    Corresponding author
    1. Department of Biotechnology, Adam Mickiewicz University, Międzychodzka 5, 60-371 Poznań, Poland, and
    2. Institute of Plant Genetics, Polish Academy of Sciences, Strzeszyńska 34, 60-479 Poznań, Poland
      *(fax +48 61 829 2730; e-mail:
    Search for more papers by this author

*(fax +48 61 829 2730; e-mail:


Since the tetraploidization of the Arabidopsis thaliana ancestor 30–35 million years ago (Mya), a wave of chromosomal rearrangements have modified its genome architecture. The dynamics of this process is unknown, as it has so far been impossible to date individual rearrangement events. In this paper, we present evidence demonstrating that the majority of rearrangements occurred before the Arabidopsis–Brassica split 20–24 Mya, and that the segmental architecture of the A. thaliana genome is predominantly conserved in Brassica. This finding is based on the conservation of four rearrangement breakpoints analysed by fluorescence in situ hybridization (FISH) and RFLP mapping of three A. thaliana chromosomal regions. For this purpose, 95 Arabidopsis bacterial artificial chromosomes (BACs) spanning a total of 8.25 Mb and 81 genetic loci for 36 marker genes were studied in the Brassica oleracea genome. All the regions under study were triplicated in the B. oleracea genome, confirming the hypothesis of Brassica ancestral genome triplication. However, whilst one of the breakpoints was conserved at one locus, it was not at the two others. Further comparison of their organization may indicate that the evolution of the hexaploid Brassica progenitor proceeded by several events, separated in time. Genetic mapping and reprobing with rDNA allowed assignment of the regions to particular Brassica chromosomes. Based on this study of regional organization and evolution, a new insight into polyploidization/diploidization cycles is proposed.


The Arabidopsis genome-sequencing project was partly prompted by the prospect of transferring information on genome structure to closely related crop species within the Brassicaceae family. The small genome size and low level of repeated sequences makes A. thaliana a perfect model, providing an array of genetic tools for further investigation of plants with more complex genomes, especially within the genus Brassica (cruciferous oilseeds and numerous vegetables). Comparative mapping approaches that use Arabidopsis-derived probes have substantially contributed to our present knowledge of the architecture and organization of Brassica chromosomes at macro- (Kowalski et al., 1994; Lagercrantz, 1998) and micro-collinearity levels (O'Neill and Bancroft, 2000; Sadowski et al., 1996). However, an unusually high rate of chromosomal rearrangements within the family (1.3 and 6.5–9.7 rearrangements per million years for the A. thaliana and Brassica lineages, respectively; Koch and Kiefer, 2005) obscures chromosomal collinearity. This limits the ability to readily determine the evolutionary relationships between these two genera on the basis of chromosomal collinearity. On average, collinear regions of only 20 cM between A. thaliana and B. oleracea could be detected by genetic mapping (Babula et al., 2003). Other Brassica species exhibit similar levels of synteny with the Arabidopsis genome (Lagercrantz, 1998). An additional difficulty derives from the high level of segmental chromosomal duplication in the Brassica lineage that probably derives from a hexaploidization event following the Arabidopsis/Brassica split (Jackson et al., 2000; Lagercrantz, 1998; O'Neill and Bancroft, 2000). Thus, reconstruction of mosaics of A. thaliana-like segmental duplications on Brassica chromosomes using RFLP mapping is considerably limited.

For this reason, there is a need to complement genetic mapping with physical mapping strategies for the analysis of Brassica genomes. Although shotgun-sequencing projects provide a large number of relatively short sequences, these cannot be aligned and thus cannot be used to reconstruct their original chromosomal position (Ayele et al., 2005). As a result, the information so obtained cannot be directly used for investigation of segmental duplications that contribute to the whole-genome architecture. Likewise, sequencing of selected BAC clones provides valuable information within individual chromosomal regions, but not on large-scale genome organization (Gao et al., 2004; Quiros et al., 2001). The construction of comprehensive BAC-based physical maps (O'Neill and Bancroft, 2000; Rana et al., 2004) is a long-term project, especially for the highly duplicated Brassica genomes. For these reasons, the FISH technique is our method of choice for investigation of chromosome segmental structure, including breakpoint organization. Unfortunately, genomic in situ hybridization (GISH), which has successfully been used for the analysis of repetitive sequence-rich grass genomes (Kosmala et al., 2006; Maestra et al., 2002, for example), cannot be utilized for our purposes. Although it has occasionally been used to study Brassica amphidiploids (Maluszynska and Hasterok, 2005; Wang et al., 2005), GISH is not suitable for tracing Arabidopsis genomic regions on Brassica chromosomes. In addition, comparative chromosome painting with probes derived from chromosome-specific DNA libraries, which has been widely used in mammals (i.e. Zoo-FISH; see Froenicke, 2005, for review), cannot be easily applied within the Brassicaceae. The chromosomes in this family are too small and insufficiently differentiated for flow-sorting. We have therefore used pools of A. thaliana BACs representing selected chromosomal regions. Such an approach was previously successfully adopted for an analysis of large chromosomal segments, or even whole chromosomes within the Brassicaceae (Lysak et al., 2001, 2003, 2005).

It has been widely accepted that the Arabidopsis–Brassica common progenitor underwent at least two whole-genome duplications, resulting in a large proportion of the genome being duplicated and lacking overlapping segments (Blanc et al., 2000, 2003; Ku et al., 2000; Ziolkowski et al., 2003). Although the most recent duplication event took place some 30–35 Mya (Ermolaeva et al., 2003), its remnants are still recognizable at the molecular level. In fact, the contemporary Arabidopsis genome is prevalently composed of up to 90 duplicated chromosomal segments of different sizes, with about 25% of the genes being conserved (Blanc et al., 2003). The mosaic structure of the genome appears to be due to large reciprocal translocations and major inversions, as has been inferred from several different genetic studies (Fransz et al., 1998; Koch and Kiefer, 2005; Lysak et al., 2003, 2005). However, the process of diploidization is still poorly understood, mainly because of difficulties in dating individual rearrangement events.

In this paper, we present an experimental approach based on comparative FISH mapping of inter-segmental breakpoints that enables us to determine the time at which the rearrangement occurred. We assume that the divergence of Arabidopsis and Brassica took place some 20–24 Mya (Koch et al., 2000, 2001; Lysak et al., 2005), relatively soon after the ancestral genome duplication. If the diploidization process ran at an extremely high rate immediately following the duplication event, the chromosomal breakpoints between individual segments should be conserved and common for both genera. In general, breakpoints that arose through rearrangements were identified in the A. thaliana genome simply by the estimation of boundaries of duplicated regions. A breakpoint-containing region (BCR) is defined as a genomic fragment between the last conserved gene from one duplicated chromosomal segment and the first conserved gene from the following duplicated segment.

Here, we report on the analysis of five BCRs in the B. oleracea genome. For this purpose, contiguous Arabidopsis BAC clones (contigs), spanning 95 BACs and 8.25 Mb in total and representing three large chromosomal regions, have been used in comparative mapping as FISH probes. The analysis can be considered as representative for the Arabidopsis genome, as the selected regions came from three distinct chromosomes at different distances from the centromeres (Figure 1). Two of them (from chromosomes 1 and 2) have been previously analysed in detail by a number of bioinformatic approaches (Ziolkowski et al., 2003). They were of special interest, as they encompass two duplicated segments derived from the last whole-genome duplication and differ markedly in size. The third region (from chromosome 3) was chosen to represent one of the largest duplicated segments in the Arabidopsis genome, and gave us the opportunity to investigate structural conservation of a region that has not significantly changed in the Arabidopsis lineage following the last whole-genome duplication. To verify and complement the FISH-based analysis, genetic mapping in Brassica was carried out using 36 expressed sequence tags (ESTs) or full-length cDNA probes from selected regions. This enabled the integration of physical and genetic maps and provided additional supporting evidence relating to some structural aspects of the chromosomal composition. A similar approach aiming at integration of genetic and physical maps has recently been presented (Howell et al., 2002, 2005).

Figure 1.

 A map of Arabidopsis thaliana chromosomes presenting regions analysed in this study.
Individual chromosomes are depicted as vertical bars, with centromeres marked as open circles. Regions used for comparative FISH mapping are shown as dark-grey and black blocks (corresponding to green and red signals in FISH experiments, respectively). Light-grey shaded bands connect corresponding duplicated segments between individual chromosomes according to the map of the Arabidopsis Genome Initiative (2000) and Blanc et al. (2003). The boundary between two segments on a chromosome defines the breakpoint-containing region (BCR). Positions of individual BCRs are indicated by arrows.


To achieve precise structural analysis and identify conservative or rearranged chromosomal segments, Brassica pachytene bivalents were physically mapped. Contigs of 15, 18 and 62 BAC clones for large chromosomal regions of A. thaliana were applied as probes. The BAC subsets of each contig were labelled with either digoxigenin or biotin conjugates, which enabled differential fluorochrome mapping. The objectives of the FISH mapping were to: (i) localize four chromosomal breakpoints contained within two A. thaliana BAC contigs in the B. oleracea genome, and (ii) study the organization of three Arabidopsis genomic regions within Brassica chromosomes.

Physical mapping of rearrangement breakpoints

In a previous paper, we analysed the relationships between duplicated chromosomal segments of Arabidopsis chromosomes 1, 2, 4 and 5 (Ziolkowski et al., 2003). These data enabled us to locate rearrangement breakpoints that took place during the diploidization of the Arabidopsis/Brassica tetraploid ancestor. Based on the bioinformatic analysis, we selected two tiling paths of BAC clones, each spanning over two BCRs (1North and 1South for chromosome 1, 2North and 2South for chromosome 2; Figure 1). Generally, for each BCR analysed, BACs representing one segment (i.e. located at one side of the BCR) were labelled with digoxigenin, while BACs assigned to the adjacent segment (i.e. located at the other side of the same BCR) were labelled with biotin. In FISH, digoxigenin- and biotin-labelled probes were detected as green and red signals respectively. Thus, the contigs belonging to these tiling paths were labelled in red–green–red manner, facilitating the analysis of both BCRs in a single experiment. An additional, fifth BCR (3North) is described below, together with a large segment from the end of chromosome 3, for which we studied the internal conservation between the species. Detailed information on the location and labelling of individual contigs on Arabidopsis chromosomes is given in Figure 1 and in Tables S1 and S2.

Heterologous hybridization with the contig from A. thaliana chromosome 2 revealed three copies of this region on B. oleracea chromosomes (Figure 2b,c), although they were unique in the model species. All three copies exhibited a similar, Arabidopsis-like signal sequence (red–green–red; Figure 2b), which clearly indicated that the entire region with two chromosomal breakpoints was conserved in the genomes of Arabidopsis (as a unique region) and Brassica (as triplicated regions). Reprobing the slides with 5S rDNA and 45S rDNA permitted determination of the chromosomal location of the corresponding regions. The two regions mapped to both arms of B. oleracea chromosome O4 (chromosome numbering according to the map of Howell et al., 2002), whereas the third region is most likely located on chromosome O3. FISH of pachytene bivalents revealed all three copies to be located proximal to the telomeres. Although individual regions differed in size on some specimens, they were of a similar size on others. For this reason, it was difficult to distinguish them clearly. The identification of chromosomes that included selected regions was further confirmed by RFLP mapping (see below).

Figure 2.

 Comparative FISH mapping using BAC contigs.
Arabidopsis BAC contigs were divided into pools of 3–11 BACs, each of which were subsequently visualized with Cy3 (red) and fluorescein (green; a, b, d, e, g). For (a) and (b), a boundary between adjacent red and green pools corresponds to a BCR.
(a) Mapping of BCR 1North on Brassica oleracea pachytene chromosomes; arrows indicate three sites with neighbouring green–red signals illustrating conservation of the regions analysed. The additional three or four red signals are probably due to non-specific hybridization of some repetitive sequences, as one of them overlaps with the major 5S rDNA locus (S).
(b, c) Mapping of BCRs 2North and 2South on B. oleracea chromosomes.
(b) Hybridization on pachytene bivalents.
(c) Localization of the region analysed on metaphase chromosomes (red, BAC probes; green, 5S rDNA). Chromosome numbers are shown for those containing the mapped regions.
(d–g) FISH mapping of the 62 BAC contigs from Arabidopsis chromosome 3.
(d) Hybridization on B. oleracea pachytene bivalents (blue, 5S rDNA as indicated by arrows; violet, 45S rDNA). Arrowheads indicate complexes 3a, 3b, 3c, 3b′ and 3c′. The BCR 3North is conserved within complex 3a, constituting the boundary between the last red and penultimate green blocks, but is not conserved within segments 3b and 3c, as red blocks 3b′ and 3c′ are separated.
(e) BAC clone hybridization (red–yellow signals) to a B. oleracea metaphase spread. Individual complexes are indicated.
(f) Reprobing of the specimen with 5S rDNA and 45S rDNA (blue and red, respectively). Metaphase chromosomes carrying the 3a, 3b and 3c complexes are numbered.
(g) Control hybridization to pachytene Arabidopsis bivalents. The scale bar indicates 20 μm.
(h) Drawing of a fragment of chromosome O4 from (d) with complex 3c and loci for 5S rDNA and 45S rDNA [signal denotations as for (d)]. The arrow indicates the position of the centromere and the arrowhead indicates the 45S rDNA locus.

More complex patterns of hybridization were obtained in experiments with the contig derived from A. thaliana chromosome 1. Although this region is unique in Arabidopsis, hybridizations to Brassica chromosomes revealed many loci. To resolve more precisely the organization of this region within the B. oleracea genome, we repeated the experiments using probes for 1North and 1South breakpoints separately. The FISH with 1South probes did not provide any evident signals, despite repeating the experiments. In contrast, hybridization with contig 1North produced a number of signals at various chromosomal positions (Figure 2a). The strongest signals resided at the heterochromatic knob of B. oleracea chromosome O4 (see S on Figure 2a), where 5S rRNA genes had previously been mapped (Ziolkowski and Sadowski, 2002). Additional signals (five or six red signals and three green signals) appeared to be relatively weak. However, all the green signals were co-located on the Brassica chromosomes with the red signals, showing partial overlap (yellow). This indicated that breakpoint 1North was conserved between the Arabidopsis and Brassica genera.

Physical mapping of the tiling path from the south arm of chromosome 3

In order to investigate the conservation of a chromosomal segment between the Arabidopsis and Brassica genera, we selected a region from chromosome 3 (Figure 1). This segment, together with its homeologue (i.e. partially homologous chromosomal region) from chromosome 2, constitutes the largest pair of duplicated segments within the A. thaliana genome, with 306 genes being conserved (; Blanc et al., 2003). The segment from chromosome 3 spans over 3.9 Mb and 947 protein-encoding genes (data from In addition, the tiling path was lengthened by 15 BACs, which enabled mapping of a breakpoint separating this segment from the rest of the chromosome. The BAC contig for the entire region (62 BACs and 5.4 Mb) was divided into six pools (labelled alternately in red–green, see Figure 1 and Table S2), which improved mapping resolution and facilitated tracing individual fragments in hybridizations to B. oleracea pachytene bivalents.

Hybridization of the tiling path to A. thaliana chromosomes confirmed that this region could be detected only as a single copy in the Arabidopsis genome (Figure 2g). It should be noted that the ancient duplication event, which took place in the common progenitor of the Brassicaceae family, could not be detected by standard FISH analysis due to the extensive sequence divergence (Lysak et al., 2001). In contrast, FISH mapping of B. oleracea pachytene bivalents revealed five corresponding fluorescent foci (Figure 2d–f). Only one of these complexes remained without any detectable changes in the Brassica lineage, with all the six blocks being conserved (complex 3a in Figure 2d). The complete conservation of this complex indicates that the chromosomal breakpoint encompassed within the region was preserved in both A. thaliana and B. oleracea. Complexes 3b and 3c comprised part of block B and entire blocks C, D, E and F (for block numbering see Table S2). Complexes 3b′ and 3c′ constituted the rest of complexes 3b and 3c, and comprised block A and part of B. In contrast to A. thaliana, in B. oleracea, all the complexes do not reside at the end of the chromosomes, but in the interstitial regions of the chromosomal arms. In addition to five main complexes, occasionally weak signals were sporadically observed, which correspond to short fragments of the regions under study.

Reprobing of the microscope slides with 5S rDNA and 45S rDNA sequences, together with genetic mapping, permitted identification of the chromosomes on which complexes 3a, 3b and 3c were localized (Figure 2f). Complex 3a and 3c were placed on chromosome O8 and O4, respectively. The third complex (3b) was located on one of the sub-telocentric non-rDNA chromosomes, most probably on chromosome O6. This result was further confirmed with RFLP mapping. Although the signals from complexes 3b′ and 3c′ were always seen at pachytene stage, they appeared only occasionally on metaphase preparations. For this reason, we were not able to assign them to individual chromosomes.

The analysis of hybridization to pachytene bivalents shed more light on the chromosomal organization of each complex. Complex 3c was localized relatively near the centromere of chromosome O4 and was flanked by the two B. oleracea 5S rDNA loci M and W (Figure 2d–f). The third locus of 5S rRNA genes (locus S; Ziolkowski and Sadowski, 2002) was localized in the centre of block C, within complex 3c. The complex is oriented with block B the most distant from the centromere. Complexes 3a and 3b were placed in the southern interstitial region, a third of the chromosomal arm length from the centromere. Complex 3a has block A most distant from the centromere, and in 3b the orientation is reversed. On average, complexes 3a, 3b and 3c were 24.0, 16.0 and 19.0 μm in length, while their counterpart in A. thaliana was 17.1 μm. A summary of the chromosomal location of all the complexes is shown in Figure 3.

Figure 3.

 Summary of physical and genetic mapping of regions from Arabidopsis thaliana chromosomes 2 and 3.
Thin vertical bars represent idiograms of A. thaliana and Brassica oleracea pachytene chromosomes. Violet and blue circles represent localization of 45S rDNA and 5S rDNA loci, respectively; individual 5S rDNA loci on B. oleracea chromosome O4 are also marked with M, S, and W as described by Ziolkowski and Sadowski (2002). Centromeres are marked by ‘Cen’. Red and green boxes represent individual blocks for segments from Arabidopsis chromosome 2 and 3 (colours as in FISH experiments). Individual conserved BCRs are marked with arrows, while non-conserved BCRs are indicated by open arrows. Long-tailed arrows connect A. thaliana regions from chromosomes 2 and 3 with their Brassica homeologues 2a, 2b, 2c and 3a, 3b, 3c respectively. Vertical bars on the right represent B. oleracea linkage groups corresponding to individual chromosomes. Horizontal lines connecting idiograms of chromosomes with linkage groups represent genes from the A. thaliana regions analysed that were used for genetic mapping in B. oleracea. Loci that were also mapped in the Brassica AG population (Howell et al., 2002; Sebastian et al., 2000) in order to assign linkage groups to equivalent chromosomes are marked in red.

Genetic mapping of regions from A. thaliana chromosomes 2 and 3

In addition to FISH, EST and full-length cDNA clones for genes from the regions under study were used for RFLP linkage analysis. Of the 57 A. thaliana clones tested, 36 were found to be informative RFLP probes (10 and 26 for regions from chromosomes 2 and 3 respectively). In total, 81 new loci corresponding to these probes were added to the existing genetic map of B. oleracea (Babula et al., 2003). Although probes for unique or low-copy genes were chosen based on sequence comparison to the whole A. thaliana genome (BLASTN, E < 10−10), eight of them were found to be homologous to a larger number of sequences both in the Brassica and Arabidopsis genomes, and these were excluded from further analysis. Finally, 45 loci out of the 81 mapped within B. oleracea genomic regions that were found to be homeologous to their original A. thaliana locations (13 and 32 loci for regions from chromosomes 2 and 3 respectively). The other loci remained unlinked or mapped to non-homeologous positions in the B. oleracea genome. For details of probes used in RFLP mapping, see Table S3.

In addition, several RFLP probes were simultaneously mapped using the reference AG population [the A12DHd (A) × GDDH33 (G) doubled-haploid mapping population], which was previously used to integrate genetic and cytogenetic maps of B. oleracea (Howell et al., 2002). This permitted assignment of our linkage groups to individual chromosomes, and finally confirmed FISH-based chromosome identification (see red loci on Figure 3).

The genetic mapping of genes from the Arabidopsis chromosome 2 region resulted in identification of three complexes located on two linkage groups corresponding to the Brassica chromosomes O3 (complex 2a, five loci) and O4 (complex 2b, three loci; complex 2c, five loci). With respect to genes from the Arabidopsis chromosome 3 region, genetic mapping also revealed three complexes located on chromosomes O4, O8 and O6. Only complex 3a located on chromosome O8 appeared to contain genes from block A, confirming the results obtained by the FISH-based mapping. The linkage groups with regions under study are shown in Figure 3. Complex 3a was defined by 15 loci and was found to be fully conserved with its Arabidopsis homeologue. Complex 3b consisted of nine loci (six in the conserved order) and complex 3c of eight loci (six in the conserved order).

One of the aims of the genetic mapping was to define more precisely the breakpoint-containing region 3North (within block B, see Table S2), which was poorly described by Arabidopsis genome analysis as a 315-kb long region between the At3g51460 and At3g52360 genes. Using 12 probes chosen from within this region, we mapped 19 loci, 11 of which were located within homeologous regions on the Brassica linkage groups. As a result, we have defined this BCR more precisely as a fragment between genes At3g51460 and At3g51820 (see Figure 4). Although we used nine probes corresponding to block A, we identified only five loci, of which four mapped to the linkage group corresponding to chromosome O8, with the other one remaining unlinked.

Figure 4.

 Comparative genetic mapping of BCR 3North.
For full genetic mapping results obtained in the study, see Figure 3. The Arabidopsis region of chromosome 3 is shown as a vertical bar with the physical position of genes as horizontal bars. The triplicated regions in Brassica oleracea are indicated on the left as thin vertical bars with chromosome designation underneath, and genetic distances between homeologues are indicated in cM. bBCR, bioinformatically defined breakpoint-containing region; gBCR, genetically limited breakpoint-containing region.


It has been accepted that the unusual segmental duplication of the Arabidopsis genome is due to breakage and reshuffling of chromosomal segments that followed the whole-genome duplication, the process known as chromosomal diploidization. The genetic and evolutionary consequences of the diploidization process include switching the duplicated genome from the tetrasomic to the disomic model of inheritance, with further modification of primary homeologous chromosomes into highly divergent ones. Although our present understanding of diploidization mechanisms is still relatively rudimentary (Wolfe, 2001), recent applications of molecular genetic techniques have revealed that it involves a number of phenomena, of which large-scale rearrangement events is one. Most of the data regarding genome changes following polyploid formation have been obtained from analyses of a few generations in newly formed allopolyploids. However, such data can provide us with only limited information, because of the quite different time scale compared to that involved in the natural diploidization process. Even so, the reports are often contradictory depending on the source of germplasm and methods applied. For Brassica species, Song et al. (1995) reported rapid genome changes in synthetic allopolyploids of Brassica juncea and Brassica napus, while Axelsson et al. (2000) suggested a conserved chromosomal organization of synthetic B. juncea. Similarly, comparisons of natural Brassica allopolyploids also resulted in contradictory conclusions [compare Cheung et al. (1997) and Axelsson et al. (2000)]. Moreover, all such genome changes have been observed for Brassica allotetraploids, with a simple addition of maternal and paternal chromosome numbers, and no evident tendency towards diploidization has been observed.

As already mentioned, the most recent whole-genome duplication of the Arabidopsis/Brassica common ancestor was followed by a diploidization process. The resulting mosaic structure of the genome has been found both for Arabidopsis (Blanc et al., 2000) and Brassica (Babula et al., 2003; Lan et al., 2000; Parkin et al., 2005). To date, no comprehensive evidence has been reported for similarity of the segmental organization along many chromosomal regions in both genera. Recently, Lysak et al. (2005), in their elegant paper, analysed the structure of a large Arabidopsis chromosomal region by heterologous FISH in a number of Brassiceae species, but the unresolved boundaries of breakpoints precluded the authors from studying their conservation in related species. In this paper, we analysed five chromosomal breakpoints corresponding to the junctions between neighbouring segments. Considering reciprocal translocations (where two non-homologous chromosomes break and exchange fragments) as accounting for the predominant fraction of genome rearrangements in the course of diploidization (Wolfe, 2001), we would expect that these breakpoints in general are likely to be created by such rearrangements. In fact, four of the breakpoints studied (1North, 2North, 2South and 3North) are well conserved in both A. thaliana and B. oleracea. The results strongly suggest that reshuffling of homeologous chromosomes accompanying the diploidization process occurred shortly after the polyploidization event. Recently, Ermolaeva et al. (2003) have demonstrated that much of the gene loss occurred separately in Arabidopsis and Brassica, suggesting that the lineages diverged shortly after ancestral genome polyploidization. Our observations are in good agreement with their investigation of chromosomal rearrangements, although the two studies addressed different levels of genome evolution.

In contrast, Arabidopsis breakpoint 3North, which is placed within block B, is conserved with respect to only one of three Brassica homeologues (Figure 3). By comparing the three B. oleracea regions, it can be noticed that two of them, i.e. 3b and 3c, differ from 3a and the Arabidopsis homeologue in the absence of block A. There are two possible scenarios explaining these structural modifications. Firstly, block A could have been attached to the ancestral segment that consisted of blocks B–F in a lineage that led to Arabidopsis and Brassica II ancestors (Figure 5). Alternatively, block A could have been separated from the ancestral segment that consisted of blocks A–F in a lineage that led to Brassica ancestor I (Figure 5). Assuming the latter scenario, the separation site of block A coincided with the ancient breakpoint 3North that took place in the Brassicaceae common ancestor (XX′ genome formation in Figure 5). Given the lack of evidence for recombination hot-spots in the Arabidopsis genome (Wright et al., 2003), the first explanation seems more probable. In support of the first scenario, it should be noted that recently published comparative linkage maps of Arabidopsis lyrata and Capsela rubella with A. thaliana suggest conservation of the 3North segment among these species (Koch and Kiefer, 2005; Yogeeswaran et al., 2005). Nevertheless, sequence data for all three homologues of an Arabidopsis gene set should be compared by phylogenetic analysis in order to confirm the first hypothesis above.

Figure 5.

 Model of diploid Brassica genome evolution via hexaploidization.
As the timescale of particular events is unknown, the lengths of the arrows do not indicate evolutionary distance but only evolutionary relations.

Irrespective of the above, our results indicate that the formation of an ancient hexaploid Brassica progenitor involved a multi-stage process, of which the final event was allo-polyploidization. This involved hybridization between genomes of the diploid (ancestor II, donor of complex 3a) and tetraploid species (ancestor I, donor of complexes 3b and 3c). Significantly, the diploid ancestor I had an Arabidopsis-like chromosome structure (organization of 3a complex, stronger hybridization signal), whereas tetraploid ancestor II showed a more diverged chromosomal composition. A monophyletic origin of the tribe Brassiceae is widely accepted (Warwick and Black, 1993, 1997; Warwick and Sauder, 2005), and, as suggested by Lysak et al. (2005), this is likely to be derived from an ancestral genome triplication. The results obtained here are expected to apply to other species within the tribe. In addition, genetic mapping of the region from Arabidopsis chromosome 3 confirmed the hypothesis that the Brassica genome had evolved via distinct steps, separated in time. Complex 3a appears to be the most closely related to its Arabidopsis homeologue, as (i) it is the only one containing block A in B. oleracea, (ii) it has the higher number of loci mapped, and (iii) it shows a complete conservation of block order (Figure 3). In contrast, both complexes 3b and 3c have been localized with a smaller number of loci and a rearranged order. This could have arisen from small-range translocations as have been observed elsewhere (Rana et al., 2004; Ziolkowski et al., 2003), but which could not be detected with FISH analysis. However, some inconsistencies in genetic mapping resulting from limitations of the technique could not be ruled out as a reason for the gene translocations detected. Overall, of the six complexes mapped in the B. oleracea genome (2a, 2b, 2c, 3a, 3b and 3c on Figure 3), four displayed micro-collinear local translocations, as demonstrated by RFLP analysis (crossing lines corresponding to genes mapped on regions 2a, 2c, 3b, 3c on Figure 3).

Our investigations also give some more general clues as to the structural evolution of plant genomes. Only 56% of Arabidopsis genes mapped in Brassica formed homeologous segments. The others were unlinked or randomly distributed along the Brassica genome. This result is in good agreement with the analyses of O'Neill and Bancroft (2000) and Rana et al. (2004), who found that half of the Arabidopsis genes were conserved in Brassica at homeologous positions, and with comparisons of genomic sequences based on B. oleracea BAC clones (Gao et al., 2004, 2005). This shows that gene loss proceeded at a higher rate in the Brassica lineage following the Arabidopsis/Brassica split. In contrast, pachytene FISH mapping of the region from A. thaliana chromosome 3 revealed that three homeologous regions in B. oleracea are larger than their counterpart in A. thaliana. Although differential chromosome condensation for both genera could be evoked, it seems unlikely that the extensive gene loss or gene translocations significantly influenced the length of the Brassica complexes. Alternatively, this might be explained by larger intergenic regions or an expansion of transposable elements in the Brassica genome (Zhang and Wessler, 2004).

We have shown that the sequence similarity of the fifth breakpoint 1South was too low between Arabidopsis and Brassica to enable successful mapping by comparative FISH. This could be due to a high level of sequence divergence and the accumulation of small rearrangements within this region, as the segment is located close to the centromere in A. thaliana. Taking this result together with different sizes of homeologous regions 3a, 3b and 3c in the B. oleracea genome, it could be concluded that the frequency of gene loss and other processes of sequence divergence can differ dependent on chromosomal location. In addition, it should be taken into account that some small fragments of the regions analysed could be translocated to other genomic positions, as some weak additional signals were occasionally observed. On the contrary, mapping of the 5S rDNA locus within one of the segments from complex 3c indicates that evolution of the repeat sequence-rich chromosomal regions proceeds at even higher rate. This locus must have been formed relatively recently, as it is not observed in complexes 3a and 3b in B. oleracea, nor in its counterpart in A. thaliana. Rapid evolution of ribosomal RNA genes is often observed when comparing the genome structure of closely related species. In summary, all the data analysed above indicate that different rates of structural evolution depend on both the chromosomal position and sequence features.

Experimental procedures

Chromosome preparation

Chromosome spreads were prepared from young buds of A. thaliana (line C24) and B. oleracea var. alboglabra (doubled-haploid line A12DHd). Pachytene chromosome preparations were prepared as described previously (Ziolkowski and Sadowski, 2002). Shortly after digestion in a solution of 0.3% w/v cellulase, 0.3% w/v cytohelicase and 10% v/v pectinase, six anthers from a single flower bud were isolated under a dissecting microscope and suspended in 60 μl of 60% acetic acid in an Eppendorf tube. Aliquots (10 μl) of the suspension were spread under a coverslip. For A. thaliana preparations, a whole flower bud was squashed on a slide in a drop of 60% acetic acid. The coverslips were removed after freezing in liquid nitrogen. Before use in FISH experiments, the slides were checked under a phase-contrast microscope.

Probe labelling

Probes for 5S rDNA and 45S rDNA were prepared as described previously (Ziolkowski and Sadowski, 2002). For comparative FISH mapping of Brassica chromosomes, the A. thaliana BACs and a single P1-derived artificial chromosome (PAC) clone were used as probes. These clones were provided by the Arabidopsis Biological Resource Center (Columbus, OH, USA). BAC DNA was isolated by the modified alkaline lysis method ( and verified by restriction analysis with EcoRI or HindIII. BAC isolates were collected into pools and labelled with biotin-dUTP or digoxigenin-dUTP, as described in Tables S1 and S2. Labelling was conducted using the standard nick-translation protocol according to the manufacturer's instructions (Roche Diagnostics, Mannheim, Germany). After labelling reactions, the probes were precipitated with ethanol and reconstituted directly in the hybridization mixture.


FISH and post-hybridization washes were performed as described previously (Ziolkowski and Sadowski, 2002) with some modifications. After fixation in 1% v/v formaldehyde in 2 × SSC, slides were briefly washed in 2 × SSC and dehydrated through an ethanol series. Subsequently, slides and probes were simultaneously denatured using a Twin-Tower device (MJ Research Inc., Watertown, MA, USA) installed on a MJ-Research 200 thermocycler (MJ Research Inc.) for 1.5 min at 80°C. The hybridization mixture of 20 μl consisted of 5–10 ng probe μl−1 for each BAC, 10% dextran sulphate, 30% formamide, 2 × SSC, 50 mm sodium phosphate pH 7.0, giving a hybridization stringency of 65%. After denaturation, the slide was incubated at 37°C in a moist chamber for 48–72 h. After hybridization, slides were washed in 2 × SSC at 37°C, then twice in 2 × SSC and 20% formamide at 37°C for 5 min each, and finally in 2 × SSC at room temperature for 5 min. The multi-colour detection of the probes was performed according to the method described by Ziolkowski and Sadowski (2002). Biotin-labelled probes were detected using Cy3-conjugated streptavidin (Sigma-Aldrich, St Louis, MO, USA). The digoxigenin-labelled probes were detected using mouse anti-digoxigenin (Roche-Diagnostics, Mannheim, Germany), anti-mouse digoxigenin (Roche-Diagnostics) and anti-digoxigenin–fluorescein (Roche-Diagnostics), sequentially.

Slides were examined using an Olympus BX-60 microscope (Olympus Optical Co. GmbH, Hamburg, Germany) equipped with a charge-coupled device (CCD) camera F-View II (Soft Imaging System GmbH, Münster, Germany). Reprobing was performed according to Heslop-Harrison et al. (1992). Images were processed by AnalySIS 3.2 (Soft Imaging System GmbH) and Adobe PhotoShop 7.0 (Adobe Systems Incorporated, San Jose, CA, USA). At least three measurements of good-quality pachytene preparations were taken for idiogram construction. However, some characteristics (e.g. chromosome arm length) were based on metaphase chromosomes.


The genetic map used in this study was generated using a set of F3 plants derived from a cross between collard (B. oleracea var. acephala) and cauliflower (B. oleracea var. botrytis) as described previously (Babula et al., 2003). A set of 57 probes used for Southern hybridization, consisting of 38 ESTs and 19 full-length clones corresponding to genes located within A. thaliana regions under study, was obtained from the Arabidopsis Biological Resource Center. The probes were chosen as representing unique or low-copy genes based on BLASTN analysis against the A. thaliana transcriptome (E < 10−10). Fifteen probes were evenly distributed along the segment from A. thaliana chromosome 2, and 42 along the segment from chromosome 3. Twelve probes were selected specifically to saturate the BCR between the A and B fragments of the latter. In addition, 5S rDNA and 45S rDNA probes were used for genetic mapping. For a list of probes used in genetic mapping, see Table S3.

Procedures for DNA extraction, restriction enzyme digestion and Southern hybridization were as previously described (Babula et al., 2003) with a minor modification. Probes were non-radioactively PCR-labelled with digoxigenin (Roche Diagnostics, Mannheim, Germany). Moderate-stringency wash conditions were used after hybridization. Genotype data from the F3 population were used to perform linkage analysis using JoinMap 3.0 software for Windows (Kyazma B.V., Wageningen, Netherlands) (van Ooijen and Voorrips, 2001). A minimum logarith of the odds (LOD) of 3.0 was set as a threshold to allocate marker loci into linkage groups, and a recombination fraction of 0.5 was used for linkage analysis. The Kosambi function (Kosambi, 1944) was used to order markers and to estimate interval distances. The final arrangement of loci was tested with the fixed-order option.

In order to identify B. oleracea chromosomes mapped by FISH, several probes were simultaneously mapped using the AG doubled haploid population (marked red in Figure 3). The population was previously used for cytogenetic and genetic linkage map integration (Howell et al., 2002; Sebastian et al., 2000). Seeds of the AG population were obtained from Warwick-HRI (Warwick Horticulture Research International, Wellesbourne, Warwick, UK). Mapping was carried out on a subset of 50 lines using the associated segregation data matrices kindly provided by the Multinational Brassica Genome Project (


We are grateful to Graham King for critical reading of the manuscript. We thank Zofia Szweykowska-Kulinska for the 45S rDNA-specific primers and Przemyslaw and Katarzyna Nuc for clone pU5SLL encoding the 5S rRNA gene. Marcell Salanoubat is acknowledged for providing PAC and BAC clones MAA21 and F22O6 respectively. The work was supported by grants numbers 2 P06A 035 28 and 3 P06A 036 24 from the State Committee for Scientific Research, Poland.