Haploid chromosome numbers vary among Brassicaceous species. In Arabidopsis the haploid set consists of 5 chromosomes, whereas many close relatives such as Capsella rubella have n = 8 chromosomes. Phylogenetic analyses within the tribe Arabideae suggested that base chromosome numbers lower than n = 8 are derived because base chromosome number reduction from n = 8 to n = 5 to 7 occurred several times (Koch et al., 1999).

The progenitors of the lineage leading to Arabidopsis and C. rubella diverged approximately 10 million years ago (Koch et al., 2000, 2001) and the first comparative mapping studies revealed conserved gene repertoires among these species (Acarkan et al., 2000; Rossberg et al., 2001). Thus, these two closely related species that differ in respect to base chromosome number offer an excellent opportunity to reveal patterns of chromosome evolution in Brassicaceae.

All markers used for genetic mapping in a population derived from an interspecific cross of Capsella grandiflora and C. rubella were sequenced in order to enable the characterization of each marker with respect to copy number and chromosomal map position in Arabidopsis. This information was used for a comparison between the Arabidopsis chromosome sequence maps and the Capsella linkage map. Extensive colinearity of these genomes was apparent but large- and small-scale deviations from genome colinearity were also identified and characterized to provide insight into factors of importance to chromosome evolution.

The annotated Arabidopsis genome sequence also facilitates the study of sequence evolution. Comparative sequence analysis of selected orthologous regions of Arabidopsis and C. rubella revealed a high degree of sequence identity for protein-coding sequences. In contrast, intergenic regions and introns are differently sized in both species, and overall sequence identity is generally not found (Acarkan et al., 2000; Rossberg et al., 2001). In the study presented here, a set of random short C. rubella sequences was compared with the annotated Arabidopsis genome sequence to assess which kind of sequences show conservation between these closely related species. A detailed analysis of sequence alignments revealed that aligned protein-coding sequences do differ from aligned intron or intergenic sequences with respect to the degree of sequence identity and the frequency of small insertions/deletions (indels). Thus, significant sequence similarities between these closely related species are not restricted to protein-coding exon sequences.

Table I. Conservation of sequence repertoires in Arabidopsis and C. rubella: a comparison of 137 sequences of C. rubella MboI fragments with Arabidopsis DNA sequences from both nuclear and organellar genomes

Using the program BLAST 2 Sequences (Tatusova and Madden, 1999), the 71 C. rubella MboI fragments that corresponded to single or low-copy regions in the Arabidopsis genome were aligned with the Arabidopsis sequence showing the highest overall DNA sequence similarity. With the parameters chosen, 23,200 of 30,856 bp of the C. rubella sequences (75.2%) were found in high scoring sequence pairs. This analysis was also performed with a 27-kbp long C. rubella region and the orthologous area of the Arabidopsis genome that is located on chromosome I (Rossberg et al., 2001; Table II).

Table II. Nature and characteristics of sequences conserved between the Arabidopsis and C. rubella genomes

A detailed analysis of the aligned sequences showed correspondence to annotated Arabidopsis protein-coding sequences for 54 out of the 71 C. rubella MboI fragments (76.1%; Table I). Based on the annotation of the Arabidopsis genome, 51.8% and 15.6% of the aligned sequences consisted of protein-coding exon and intron sequences, respectively. The remainder of the alignments, 32.5%, which included 5′- and 3′-untranslated regions of genes, was classified as intergenic sequences (data not shown). The comparison of the contiguous C. rubella genomic DNA region and its orthologous counterpart in Arabidopsis revealed that the alignments consisted of 55.7% protein-coding exon, 9.0% intron, and 35.3% intergenic sequences (data not shown). Thus, the fraction of aligned intron sequences in the contiguous orthologous regions was, at 9%, much lower than the fraction observed for the dataset of the aligned random MboI fragments (15.6%). This difference between the two datasets reflected that only about 20% of the Arabidopsis genic sequences corresponded to intron sequences in this particular region of Arabidopsis chromosome I (Rossberg et al., 2001), whereas the average value for the Arabidopsis genome amounts to approximately 35% (Arabidopsis Genome Initiative, 2000).

Sequence identity values of about 90% were found for aligned protein-coding exon sequences, whereas values of approximately 80% were observed for aligned sequences that consisted of intron or intergenic sequences regardless of whether the dataset of the random MboI fragments was analyzed or whether the contiguous orthologous regions were evaluated (Table II).

In total, 291 small indels with an average length of 3.4 bp (1–21 bp) were found in the aligned sequences of the MboI fragments. In the alignments of the contiguous orthologous regions, 254 indels were found. These indels ranged in length from 1 to 18 bp and spanned on average 3.2 bp (Table II). Regardless of which of the two datasets was analyzed, indels were observed on average once every 80 bp, but they were much more frequent in aligned intron or intergenic sequences than in exon sequences (Table II). The alignments of intron and intergenic sequences showed indels on average once every 40 bp and more than one-half of the indels spanned 1 or 2 bp. In contrast, the sizes of indels in alignments of exon sequences corresponded to one or more codons. Treating all indels as small insertions showed that this type of sequence alteration was approximately twice as frequent in the C. rubella sequences when compared to the Arabidopsis sequences (data not shown).

A single codominant polymorphism was scored for each of 84 RFLP markers, 2 for 13 markers, and 3 for a single marker in Capsella. RFLP mapping of sequences derived from the 18S-25S rDNA loci in Arabidopsis identified 2 loci in Capsella, and a codominant polymorphism was scored for locus rDNAa, whereas a dominant polymorphism was evaluated for locus rDNAb. Additionally, 19 loci were placed on the Capsella map using PCR-based methods. The resulting genetic map consisted of 133 codominant loci distributed over 8 linkage groups and spanned 582.1 cM (Fig. 1).

Figure 1. Genetic linkage map of Capsella. Linkage data for the Capsella mapping population derived from an interspecific cross were calculated with Map Manager QTX (Manly et al., 2001). Using a logarithm of odds ratio score of 5.0, 8 linkage groups labeled A to (more ...)

Three markers (E6, E9, and E20) harboring mitochondrial DNA sequences showed maternal inheritance; thus all F₂-plants carried the C. grandiflora allele of these markers. For a nuclear-encoded codominant locus, the expected segregation among the F₂ progeny is a 1:2:1 ratio of plants homozygous for the C. grandiflora allele, heterozygous, and homozygous for the C. rubella allele, respectively. The results of χ²-tests (P = 0.05) revealed that the observed segregation ratios were significantly different from the expected distribution for 25 of the 133 loci (Fig. 1). All but 3 (E35b and mi353, linkage group B; E57, linkage group D) of these 25 loci map to linkage groups F and G. Markers CL5.1 and m326 delimit a chromosomal region on linkage group G that is characterized by a significant under-representation of homozygous C. grandiflora plants. In contrast, C. grandiflora alleles are significantly over-represented for 8 of the loci showing a distorted segregation (E31, mi219, mi90, mi433, mi138, mi438, mi74b, and mi174), which map to linkage group F. Segregation distortion has previously been noted for plant populations derived from interspecific crosses (e.g. Livingstone et al., 1999).

Figure 2 shows a comparative map of the 8 Capsella linkage groups and the 5 Arabidopsis chromosomes. A comparison of the map positions of the 62 single-locus marker sequences with those of the corresponding loci in Capsella clearly indicated that 61 locus pairs reside in colinear positions on the Arabidopsis and Capsella maps. The mapping of marker E80, which represented a single-copy sequence in Arabidopsis as well as in Capsella, revealed a translocation, the extent of which is unknown.

Figure 2. Comparative map between Arabidopsis and Capsella. The sequence maps of the five Arabidopsis chromosomes are shown. Horizontal bars mark the positions of the centromeres. The designations of the chromosomes (I to V) indicate the telomeric ends of the short (more ...)

For 48 out of the 54 markers for which 2 or more loci were found in the Arabidopsis genome, either the marker sequence itself or its closest Arabidopsis homolog resided in a colinear position with the genetically mapped Capsella locus. Colinearity between a Capsella locus and a homolog of a marker sequence was found for an additional 15 locus pairs.

In total, comparative mapping revealed 124 locus pairs distributed over 14 large colinear segments on the Arabidopsis and Capsella maps. Two of these segments may harbor small-scale inversions and/or translocations (Capsella linkage group B, Arabidopsis chromosome I; Capsella linkage group H, Arabidopsis chromosome V). Each of the five Arabidopsis chromosomes corresponds to two or three different Capsella linkage groups or segments thereof. Thus, translocations or fusions of large chromosome segments were an important factor in differentiating the genomes of the progenitors of Capsella and Arabidopsis since their divergence. Nuclear organizer regions (NORs) adjoin the telomeres on the short arms of Arabidopsis chromosomes II and IV (Copenhaver and Pikaard, 1996). RFLP mapping of the 18S rDNA sequences in Capsella also revealed two rDNA loci, but neither of these two loci map in a colinear position with the rDNA loci in Arabidopsis, indicating that the NORs had also been involved in translocations. Furthermore, evidence for inversions of large chromosome segments was found (Capsella linkage group B, Arabidopsis chromosome I; Capsella linkage group F, Arabidopsis chromosome IV; Capsella linkage group H, Arabidopsis chromosome V).

For each of five markers that mapped to 2 loci in Arabidopsis (E27, E54, E72, E76, and mi358), 2 loci were also found on the Capsella genetic map. Colinear positions on the Arabidopsis and Capsella maps were found for all 10 locus pairs. With the exception of marker mi358, these markers resided in duplicated regions of the Arabidopsis genome. Likewise, the positions of loci E92a, E92b, mi330a, and mi330b in the Capsella genome were found to be colinear with the Arabidopsis loci, mapping to a duplicated region between Arabidopsis chromosomes IV and V (Fig. 3A). These results are consistent with the occurrence of this segmental duplication in the Capsella genome.

Figure 3. Comparative mapping of segmental duplications. Arabidopsis markers mapping to duplicated regions of the Arabidopsis genome have been used for genetic mapping in Capsella. Selected parts of a segmental duplication harboring duplicated sequences of markers (more ...)

In addition to large-scale rearrangements, evidence for small-scale deviations from colinearity was found. For example, Capsella loci C57, E92c, and mi320a did not have a corresponding Arabidopsis sequence in a colinear position. In three cases copy-number changes were detected; for Arabidopsis single-locus markers E82, m315A, and mi74, two loci each were mapped in Capsella. Despite the fact that markers mi74 and E82 are single-locus sequences in Arabidopsis, they are located in segmentally duplicated regions of the Arabidopsis genome. The comparative mapping results indicate that two copies of markers mi74 and E82 each should have been present in the duplicated segments harbored by the common progenitor of Arabidopsis and Capsella, but one copy for each of the markers was lost in the lineage leading to Arabidopsis, whereas both copies of markers mi74 and E82 were retained in the lineage leading to Capsella (Fig. 3B and data not shown).

Two loci were found in Capsella for Arabidopsis single-locus marker mi335. Because marker mi335 harbors three different protein sequences (Supplemental Table S1), it cannot be discriminated whether this deviation from colinearity is due to a copy number change or a small-scale translocation involving part of the marker sequence.

Earlier comparative sequence analyses of orthologous regions of Arabidopsis and C. rubella have revealed a high degree of sequence identity for protein-coding sequences. In contrast, intergenic regions and introns are differently sized in both species, and overall sequence identity is generally not found (Acarkan et al., 2000; Rossberg et al., 2001). The study of the alignments between the Arabidopsis and C. rubella sequences presented here corroborates these findings. The Arabidopsis genome consists of approximately 28.8% protein-coding exons, 15.7% introns, and 55.5% intergenic sequences (Arabidopsis Genome Initiative, 2000). Alignment of C. rubella DNA sequences with the corresponding Arabidopsis sequences revealed that protein-coding sequences represented approximately one-half of the aligned sequences, whereas the other one-half consisted of intron and intergenic sequences (Table II). Thus, significant levels of sequence conservation in the alignments are over-represented in protein-coding exon sequences when compared to intron and intergenic sequences. Nevertheless, it is important to note that significant levels of sequence conservation between Arabidopsis and Capsella are not restricted to protein-coding exon sequences.

Regardless of whether sequences of random C. rubella MboI fragments or of a contiguous 27-kbp region were compared to the corresponding sequences in Arabidopsis, the average degree of sequence identity was approximately 10% higher in aligned protein-coding sequences than in that of aligned intron or intergenic sequences. Comparative sequence analysis of eight orthologous genes in Arabidopsis and C. rubella revealed approximately 90% sequence identity at the nucleotide level, but for a rapidly evolving gene a much lower value of about 80% was observed (Rossberg et al., 2001). Thus, the observed 10% difference in average sequence identity levels in aligned exon sequences compared to alignments of intron or intergenic sequences is not sufficient to distinguish aligned protein-coding sequences from alignments of intron or intergenic regions in an unambiguous manner. In contrast, both indel size and frequency are appropriate features to differentiate between aligned protein-coding exon sequences and alignments of intron or intergenic sequences of closely related species. Detailed inspection of the alignments revealed that these features are especially powerful to help determine the beginning and end of an open reading frame. Thus, the current weakness of gene structure prediction programs to identify coding region limits (Mathé et al., 2002) may be overcome by taking advantage of data concerning indel size and frequency in aligned sequences of closely related species.

The analysis of double strand break repair revealed that larger deletions were found more frequently in Arabidopsis than in Nicotiana tabacum. Whereas 40% of the deletions were accompanied by insertions in N. tabacum, this was not the case in Arabidopsis. Based on these results, Kirik et al. (2000) proposed that species-specific differences in double strand break repair might influence genome evolution. Interestingly, through analysis of indel frequencies in the alignments between the Arabidopsis and C. rubella sequences, another species-specific difference concerning sequence evolution was revealed. Treating all indels in the alignments of the Arabidopsis and C. rubella sequences as insertions, it emerged that small insertions occurred approximately twice as often in C. rubella sequences than in Arabidopsis sequences (data not shown). However, it was not possible to discriminate whether these results reflect an overall higher rate of small insertions in C. rubella and/or a generally higher rate of small deletions in Arabidopsis.

In comparative genetic mapping experiments, it is often not possible to distinguish deviations from colinearity from the mapping of paralogous sequences (Bennetzen, 2000). This shortcoming was overcome in the study of the Arabidopsis and Capsella genomes due to the availability of the Arabidopsis chromosome sequence maps (Arabidopsis Genome Initiative, 2000), which allow the establishment of copy numbers and positions for all markers used in the Arabidopsis genome. For example, marker E92 has 10 copies in the Arabidopsis genome (data not shown). In Capsella, three polymorphic loci and several monomorphic fragments were detected (data not shown). Loci flanking the three different loci on the Capsella linkage maps were found in colinear arrangements in Arabidopsis. In the colinear blocks defined by the flanking markers orthologous locus pairs were found for E92a and E92b, whereas none of the 10 copies corresponding to marker E92 was present in a colinear arrangement with E92c (Figs. 2 and 3A). Thus, mapping the E92c locus revealed a deviation from colinearity.

The availability of the Arabidopsis chromosome sequence maps was also exploited to ensure good coverage of the comparative map because it offers the opportunity to target any particular region of the genome for a comparative mapping study. This is particularly useful if mapping results indicate a deviation from colinearity. For example, the order of markers mi330b and mi194 on Capsella linkage group H was inverted when compared to their arrangement on the sequence map of Arabidopsis chromosome V. Mapping of additional markers (E54a, E92a, and mi61) located in this interval of the Arabidopsis genome unambiguously showed that this was due to a large chromosomal inversion and not by a translocation of a chromosomal segment harboring a marker sequence (Fig. 2).

Comparative physical mapping in Arabidopsis and C. rubella revealed that one particular region duplicated between Arabidopsis chromosomes I and III was also found in two copies in C. rubella (Rossberg et al., 2001). The comparative study of several markers located in duplicated regions of the Arabidopsis genome that were investigated here gave comparable results (Fig. 3 and data not shown), indicating that these segments should have been present in the common ancestor of Arabidopsis and Capsella (Fig. 3 and data not shown). Consistent with this finding the estimated values for the age of the duplication events in the Arabidopsis genome (Lynch and Conery, 2000; Vision et al., 2000; Bowers et al., 2003; Ermolaeva et al., 2003; Raes et al., 2003) are much higher than the divergence time of Arabidopsis and Capsella, which was calculated at 10 million years ago (Koch et al., 2000, 2001). Gene loss in duplicated regions of the genome has been proposed to be an important factor in shaping plant genomes (for review, see Schmidt, 2002); the comparative mapping results for markers E82 and mi74 lend further support to this view (Fig. 3B and data not shown).

Comparing the organization of the colinear blocks and the NORs in the two genomes unveiled 14 large chromosomal rearrangements. In addition to these changes involving large chromosome segments, the genome arrangement of the two species is distinguished by numerous small rearrangements. Approximately 6% of the analyzed loci revealed such changes, which included deletions/insertions, duplications, and/or translocations of gene sequences. No attempt was made to map all loci corresponding to the different marker sequences in Capsella; thus it is reasonable to assume that such small-scale changes are far more frequent than indicated by the data presented here.

Koch et al. (1999) concluded from results of phylogenetic studies that base chromosome numbers lower than n = 8 are derived in the tribe Arabideae. The results of comparative mapping between Arabidopsis and Capsella showed that such a reduction of base chromosome number cannot be attributed exclusively to chromosome fusions. If the chromosome number n = 8 of Capsella represents the ancestral state, at least 6 chromosome fusion (involving linkage groups A/B, C/D, C/E, G/F, and F/G/H) and three chromosome fission events (involving linkage groups C, F, and G) must have taken place for the observed arrangement of the 14 colinear blocks in the 8 Capsella and the 5 Arabidopsis chromosomes to have occurred. Furthermore, comparative mapping revealed that the 2 genomes are distinguished by three large chromosomal inversions and 2 translocations involving the NORs.

Previous comparative genome analyses in the Brassicaceae have largely been focused on the different Brassica species or on comparisons of these genomes to that of Arabidopsis (for review, see Schmidt et al., 2001). Comparative mapping using Brassica oleracea, Brassica rapa, and Brassica nigra revealed an almost complete conservation of gene repertoire, but the genomes were distinguished by multiple rearrangements (Lagercrantz and Lydiate, 1996). Analyses between Arabidopsis and B. oleracea led to the detection of conserved linkage arrangements (Kowalski et al., 1994; Lan et al., 2000; Babula et al., 2003; Lukens et al., 2003), but these were much smaller than the sizes of the colinear blocks that were observed for the Arabidopsis and Capsella genomes. These encompassed 7.2 Mbp of the Arabidopsis genome and 38.2 cM of the Capsella genome. This finding may in part reflect the more ancient divergence of the Arabidopsis and Brassica lineages when compared to that of the progenitors of Arabidopsis and Capsella (Yang et al., 1999; Koch et al., 2000, 2001). However, it should also be noted that comparative mapping indicated the presence of many duplicated and triplicated segments in the B. oleracea, B. rapa, and B. nigra genomes. Detailed microcolinearity studies between selected genomic regions in Arabidopsis and B. oleracea showed that the separation of the lineages leading to Arabidopsis and Brassica predated the triplications seen in the Brassica genomes (O'Neill and Bancroft, 2000; Schmidt et al., 2003). Thus, it is conceivable that the high number of chromosomal rearrangements that distinguish the Arabidopsis from the B. nigra and B. oleracea genomes is at least in part due to the relatively recent polyploidization in the Brassica species (Lagercrantz, 1998; Lukens et al., 2003).

Probes derived from Arabidopsis chromosome IV were used for comparative chromosome painting in the closely related species Arabidopsis halleri, Arabidopsis lyrata, Cardaminopsis carpatica, and C. rubella, all of which share a base chromosome number of n = 8 (Lysak et al., 2003). Importantly, in all four species, the arrangement of colinear segments was consistent with the results of the comparative mapping between Arabidopsis and Capsella. Furthermore, the homeologs of Capsella linkage group G in these species carried a NOR in the same position as indicated by the RFLP mapping studies of 18S-25S rDNA sequences in Capsella (Figs. 1 and 2). These results clearly show that the genetic linkage map of Capsella based on Arabidopsis markers that we have established will prove to be an indispensable tool for chromosome mapping studies in close relatives and the study of chromosome evolution in Brassicaceae.

C. rubella sequences were aligned with the corresponding Arabidopsis sequences using the program BLAST 2 sequences (Tatusova and Madden, 1999) with the same parameters used for the BLASTN analysis.

The Arabidopsis cDNA clones used as RFLP markers were denoted as follows: E5 (VBVEA05), E6 (FAFM25), E7 (VBVAH05), E9 (YAP234T7), E13 (139A22T7, AJ608275), E17 (172G2T7), E24 (102B12T7), E25 (198N17T7), E26 (104M9T7), E27 (92I17T7), E30 (YAY106), E31 (241C22T7), E32 (3H8TT), E35 (YAY337), E36 (85E6T7), E53 (113M1T7, AJ608276), E54 (G10A6T7), E57 (133K4T7), E61 (149C21T7), E64 (192P5T7), E65 (174E13T7), E66 (166D7T7), E71 (91P17T7), E72 (OAO172), E73 (206L7T7), E74 (109G22T7), E76 (176G19T7), E79 (VBV08–30792), E80 (TAP0180), E82 (VBVEB09), E83 (AJ608277), E92 (192F6T7), E94 (OAO217), E96 (G2F2T7), E98 (198N17T7), and E99 (c13.049, AJ299418).

Table III. Oligonucleotide combinations used for amplification of PCR-based markers from Capsella genomic DNA

For marker Cos2, a pronounced size difference between the C. rubella and C. grandiflora allele sequences was exploited for the mapping experiments. Single-strand conformational polymorphism analysis (Slabaugh et al., 1997) was carried out for loci A7, A9, CA6, CCr8, CE6, CH16, CI18, CL5.1, CL9, CL16, CM7, CO11, and Cos57. The PCR-amplified products were separated on gels prepared with MDE gel solution (FMC Bioproducts, Vallenbak Strand, Denmark) and the resulting gels were stained with silver (Sanguinetti et al., 1994). For loci CL19, CN18, Cos9, and Cos36 the PCR-amplified products were subjected to restriction with HindIII, HindIII/ClaI, HindIII/BamHI, and RsaI, respectively, prior to single-strand conformational polymorphism analysis.

An oligonucleotide combination that had been developed for the amplification of Arabidopsis sequences was used to amplify genomic DNA sequences of both Capsella species for locus A20 (5′-gcttccaaggctttgattctg-3′, 5′-ggcttagtctgaacaggttcg-3′). Restriction of the resulting PCR products with DraI revealed a polymorphism between C. rubella and C. grandiflora.

To establish the copy-number of Capsella PCR-based markers, the following parameters were used for BLASTN: nucleotide match 1, nucleotide mismatch −2, gap open penalty 5, and gap extension penalty 1. All corresponding Arabidopsis sequences that showed E < 10⁻⁹ were then mapped onto the sequence maps of the Arabidopsis chromosomes.

To reveal whether a particular marker was mapping to segmental duplications in the Arabidopsis genome the protein coding sequences harbored in both the marker sequence and its homolog(s) were assessed. The resulting gene codes and those of genes in the immediate vicinity of these genes were used to search the datasets describing the segmental duplications of the Arabidopsis genome (http://www.tigr.org/tdb/e2k1/ath1/Arabidopsis_genome_duplication.shtml; http://mips.gsf.de/proj/thal/db/gv/rv/rv_frame.html).

Sequence data from this article have been deposited with the EMBL/GenBank data libraries under accession numbers AJ581160–AJ581296 and AJ608275–AJ608277.

Prof. Dr. H. Hurka (Universität Osnabrück) kindly donated seed of C. rubella and C. grandiflora. Arabidopsis ESTs and RFLP markers were provided by the Arabidopsis Biological Resource Center (ABRC, Ohio State University). We thank the greenhouse staff at the Max Delbrück Laboratory for taking care of the plants. We thank M. McKenzie for carefully editing the manuscript and I. Witt for comments on the manuscript.