Introduction

Centromeres are the domains of chromosomes that are responsible for sister chromatid cohesion, kinetochore assembly and spindle attachment, and are essential for proper chromosome segregation during mitosis and meiosis. In most eukaryotes centromeres are embedded within long arrays of tandem DNA repeats (Csink and Henikoff 1998; Heslop-Harrison 2000; Lamb et al. 2004). The lengths of the repeat units show strikingly little variation, likely determined by the nucleosome unit length (150–180 bp) or its multiple (for review see Henikoff et al. 2001). Examples of such repeats are the 171 bp satellite in primates (Wevrick and Willard 1989), the 180 bp pAL1 repeat in Arabidopsis thaliana (Arabidopsis) (Martinez-Zapater et al. 1986; Maluszynska and Heslop-Harrison 1991) and the 155 bp CentO repeat in rice (Oryza sativa) (Cheng et al. 2002). These repeats form long arrays with a head-to-tail orientation and will be referred to here as satellite repeats, irrespective of their nucleotide composition.

The centromeric satellite repeats of different species display little or no sequence homology and even ecotype-specific sequence variants have been reported, as in the case of Arabidopsis (Hall et al. 2003). In contrast to the rapidly evolving centromeric DNA sequences, proteins associated with centromeric sequences can be highly conserved. One example is the centromere-specific histone H3, CENH3, specifically associated with the functional centromere in yeast, animals and plants (see for review: Jiang et al. 2003; Houben and Schubert 2003). But even in the case of CENH3 there is evidence of increased rapid evolution, while other centromere-associated proteins appear to be specific to certain organisms (Henikoff et al. 2001). The occurrence of satellite repeats is not restricted to centromeric regions, as they can also be located in pericentromeric heterochromatin and in non-centromeric heterochromatic regions such as heterochromatic knobs in maize (see for review Heslop-Harrison et al. 2003; Hall et al. 2004).

The legume Medicago truncatula (Medicago) has been selected as a model plant due to its close phylogenetic relationships to numerous crop legume species (Choi et al. 2004) and its well-characterized symbiosis with Rhizobium bacteria. It has a relatively small genome of 560–580 Mb per 1C and eight chromosome pairs (Agarwal and Gupta 1983; Blondon et al. 1994). Chromosome morphology is most distinct in 4′,6-diamidino-2-phenylindole (DAPI)-stained pachytene chromosomes, revealing blocks of heterochromatin in the pericentromeric and nucleolar organizer region, in contrast to the mostly euchromatic chromosome arms. We have identified the MtR1 satellite repeat by end sequencing of randomly selected bacterial artificial chromosome clones, and mapped it to six sites of pericentromeric heterochromatin (Kulikova et al. 2001). We pursued our search for other satellite repeats that are part of the functional centromere itself and the flanking heterochromatic areas. This was done by exploiting sequence data from a low-pass random shotgun survey of the Medicago accession Jemalong A17, revealing two additional satellite repeats, MtR2 and MtR3. The organization of these repeats was analyzed and their chromosomal positions established by fluorescent in situ hybridization (FISH) on pachytene and metaphase chromosomes. As repeats may contribute considerably to the variation in genome size between related taxa, as described for the rice indica and japonica cultivar groups (Ohmido et al. 2000), we compared the distribution of MtRs in Jemalong A17 with that of Medicago accession R108-1, a genotype that possesses a genome 20% smaller than that of Jemalong A17 (Blondon et al. 1994; Hoffmann et al. 1997).

Materials and methods

Plant material

Young leaves of the Medicago accessions Jemalong A17 and R108-1 were used for DNA isolation (Rogers and Bendish 1988) and Southern blot analysis. Young flower buds were used to prepare cell spreads containing pollen mother cells at meiosis (Ross et al. 1996; Kulikova et al. 2001).

DNA sequences and analysis

The procedures for random shotgun cloning, fluorescent-based DNA sequencing and subsequent analysis were described in Pan et al. (1994), Bodenteich et al. (1993) and Sambrook et al. (1989). Briefly, 50 μg portions of leaf genomic DNA of Medicago line A17 were randomly sheared for preparing blunt-end fragments. After T4 DNA polymerase, T4 kinase treatment and gel purification, fragments in the 1–3 kb range were ligated into SmaI-cut, bacterial alkaline phosphatase-treated pUC18 (Pharmacia), and Escherichia coli strain XL1BlueMRF’ (Stratagene) was transformed by electroporation. A random library of approximately 12,000 colonies from each transformation was picked to 384-well plates, grown in Terrific Broth (TB) medium supplemented with 100 μg of ampicillin for 14 h at 37°C with shaking at 250 rpm, and the sequencing templates were isolated by means of the alkaline lysate protocol.

Sequencing reactions were performed as previously described (Chissoe et al. 1995) using either the Applied Biosystems Big Dye 3.0 or Amersham ET terminator sequencing reaction mixes with the M13 universal forward and reverse primers. The reactions were incubated for 60 cycles in a Perkin-Elmer Cetus DNA Thermocycler 9600 at cycle conditions recommended by the manufacturer. Unincorporated dye terminators were removed by ethanol precipitation at 20°C, and after the fluorescent-labeled nested fragment sets had been dissolved in double-distilled water they were resolved by electrophoresis on an ABI 3700 capillary DNA sequencer. After base calling with the ABI Analysis Software, the analyzed data were transferred to a Sun Workstation Cluster and reanalyzed using Phred and assembled using Phrap (Ewing et al. 1998; Ewing and Green 1998). Overlapping sequences and contigs were analyzed using Consed (Gordon et al. 1998).

Southern and dot-blot analysis

Digested genomic DNA (3 μg/lane) was separated on a 1% agarose gel, blotted onto Hybond N+ membranes (Amersham Pharmacia Biotech) and hybridized with the 32P-labeled DNA probes. The probes were made by the polymerase chain reaction (PCR) with primers specific to each satellite repeat using plasmid DNA as template. MtR1 had been subcloned in pGEM-T (Promega) previously (Kulikova et al. 2001). A 283 bp fragment of MtR2 was amplified by PCR with two oligonucleotide primers 5′-tgactagtcaacttaggaac-3′ and 5′-tgagaaattcaagttactgc-3′, while the 169 bp MtR3 fragment was amplified with the primers 5′-tgttcattctagttatgtaatac-3′ and 5′-tatgagacatttagcaagaatc-3′. The PCR fragments were subcloned in pGEM-T. Southern and dot-blot hybridizations were carried out in phosphate–SDS buffer at 65°C for 16 h followed by washing with 2×SSC, 0.1% SDS (10 min), 1×SSC, 0.1% SDS (20 min) and 0.2×SSC, 0.1% SDS (20 min). The membranes were exposed in PhosphoImager cassettes and signals were visualized using a Storm PhosphoImager (Molecular Dynamics). Dilution series (100, 50, 10, 5, 1 and 0.1 ng) of genomic DNA and dilution series (5, 2.5, 1, 0.5, 0.25, 0.1 and 0.01 ng) of plasmid DNA containing a repeat sequence were spotted on a membrane and hybridized with the corresponding repeat. The hybridization signals were quantified with Image Analysis Software (Image Quant TL; Amersham Biosciences). Comparison of the radioactivity bound within the linear range of response was used to estimate the relative amount of the repeats per haploid genome.

Fluorescent in situ hybridization

The FISH protocol used was described in Kulikova et al. (2001). The MtR1, MtR2 and MtR3 probes were labeled by PCR with either digoxigenin-11-dUTP or biotin-16-dUTP (Roche), and clone pCT4.2, containing 5S rDNA (Campell et al. 1992), was labeled by nick translation according to the manufacturer’s protocol (Roche). Biotin-labeled probes were detected by initial binding with Cy3-conjugated streptavidin (Jackson ImmunoResearch Laboratories), followed by amplification with biotinylated anti-streptavidin (Vector Laboratories, USA) and a second treatment with streptavidin-Cy3. Digoxigenin-labeled probes were detected with sheep anti-digoxigenin-fluorescein (Roche) and amplified with rabbit anti-sheep-FITC (fluorescein isothiocyanate; Jackson Immuno-Research Laboratories). Chromosomes were counterstained with 5 μg/ml DAPI in Vectashield antifade solution (Vector Laboratories). Slides were examined under a Zeiss Axioplan 2 Photomicroscope equipped with epifluorescence illumination, and filter sets for DAPI, FITC and Cy3 fluorescence. Selected images were captured by a Photometrics Sensys 1305×1024 pixel CCD camera and processed with Genus Image Analysis Workstation software (Applied Imaging Corporation, Newcastle upon Tyne, UK). Chromosome lengths and fluorescence signals were measured with Image ProPlus 5.0 (Media Cybernetics, Silver Springs, Md.).

Results

Identification of Medicago satellite sequences

In total 23,544 sequence reads were generated from randomly picked clones, encompassing 20.4 Mb or ∼4% of the Medicago genome. Micropeats software (Parsons 1995) was used to identify three different tandem repeats with unit lengths varying from 166–183 bp. One of these repeats is similar to the previously described repeat MtR1 (166 bp) (Kulikova et al. 2001). Two new repeats, named MtR2 and MtR3, were found that have motif lengths of 183 and 166 bp, respectively. Sequence data from individual reads revealed that MtR1, MtR2 and MtR3 are tandem repeats with a typical head-to-tail orientation. Within each satellite family, several polymorphic variants were found. Therefore we determined a consensus sequence for each of the three repeat elements (Fig. 1). Initial repeat units were used to locate all identical and degenerate units in the complete data set using standard BLAST (Altschul et al. 1990, 1997). The majority-rule consensus sequences of the repeat units were determined by aligning the individual units extracted from the BLAST hits in CLUSTAL W (Thompson et al. 1994).

Fig. 1
figure 1

The consensus sequences of the Medicago satellite DNAs

Full size image

Within individual sequence reads these three repeats are not interrupted by other sequences. To analyze whether MtR1, MtR2 and MtR3 indeed form long tandem arrays, we performed Southern blot analysis of restriction enzyme-digested total genomic DNA. The restriction enzymes that were used have only a single restriction site in the repeat unit. The molecular weights of the laddered hybridizing fragments correspond to monomers and multimers of the appropriate unit size (Fig. 2a), showing that these repeats indeed form long tandem arrays. The observation of multimeric fragments may be explained by sequence heterogeneity between repeats, leading to loss of restriction enzyme sites, or due to chromatin modification (e.g., methylation) resulting in restriction enzyme resistance at certain sites.

Fig. 2a,b
figure 2

Genomic organization of MtR1, MtR2 and MtR3 repeats in Medicago truncatula. a Southern blot analysis of satellite DNAs in the Jemalong A17 genome. Typical long tandem repeat ladder patterns are visible after digestion of genomic DNA and hybridization with the corresponding MtR repeat. b MtR3 is present in the R108-1 genome and is also organized in long arrays. Genomic DNA of both accessions was digested with DraI

Full size image

Sequence comparison of the three MtR repeat units failed to detect similarity, suggesting that these repeats either arose independently or have diverged significantly since radiation from a common ancestor. To date, only two centromeric satellite repeats have been identified in legumes: the 488 bp tandem repeat pVuKB1 of Vigna unguiculata (Galasso et al. 1995) and a 120 bp repeat from Glycine max (L.) Merr. (Morgante et al. 1997). Pairwise alignments (ClustalW, version 1.82; http://www.ebi.ac.uk/clustalw) between the Medicago repeats and those of V. unguiculata or soybean failed to detect regions of sequence similarity. Furthermore, the MtR satellite DNAs lack detectable similarity to satellite repeats from other organisms.

Quantification of the Medicago satellite repeats

The frequency at which the repeats occur in the 20 Mb of random genome shotgun sequence was used to estimate the percentage of the Medicago genome that is composed of the three satellite repeats. The most abundant repeat is MtR2, constituting 4.5% of the genome, whereas that of MtR1 constitutes 1.1% and that of MtR3, 0.95%. We also determined the fraction of the Medicago genome corresponding to each MtR by means of quantitative dot-blot hybridization. Analysis of hybridization signals revealed that MtR1, MtR2 and MtR3 make up 1.4, 4.4 and 1.1%, respectively, of the Medicago genome. These values closely correspond to those obtained from the sequence data.

Mapping by FISH shows that MtR3 is located at the centromeres

We used FISH to locate the three MtR satellite repeats on Medicago pachytene chromosomes. Late pachytene microsporocytes stained with DAPI display brightly fluorescing (dark blocks in the inverted images) heterochromatic blocks in the pericentromeric regions (Fig. 3a,c,e) and a constriction of slightly weaker stained chromatin within the putative functional centromere (indicated by arrowheads in Fig. 3a). MtR1 was previously shown to be located within pericentromeric heterochromatin on the long arms of chromosome 1, 2 and 8, and on the short arms of chromosome 4 and both arms of chromosome 7 (Fig. 3d). MtR2 is located in all 16 pericentromeric heterochromatic blocks (Fig. 3b) and is typically proximal to that of MtR1, with the exception of chromosome 2 where the positions of the MtR1 and MtR2 regions are reversed (Fig. 3d). MtR3 is located in the centromeric regions (visible as primary constrictions) of all chromosomes (Fig. 3f,g). In addition to its centromeric location, the MtR3 region clearly extends into the pericentromeric regions of chromosomes 2, 3 and 5 (Fig. 5). Further evidence for the location of MtR3 at the functional centromere was obtained from FISH images of meiotic metaphase/anaphase I complements in which the fluorescence signals were located at the centromere domains of the homologous chromosomes pulled to opposite poles, suggesting that MtR3 resides at the kinetochores where spindle fibers are attached (Fig. 3h,i).

Fig. 3a–i
figure 3

Localization of MtR satellites on Jemalong A17 chromosomes by fluorescent in situ hybridization (FISH) and counterstaining with 4′,6-diamidino-2-phenylindole (DAPI). a,c,e Pachytene complements. Centromeres (indicated by arrowheads in a) are located between heterochromatic blocks and visible as primary constrictions. b,d,f FISH of pachytene complements with MtR satellite probes. b MtR2 (red) is located in pericentromeric heterochromatin of all chromosomes. d Localization of MtR1 (green) and MtR2 (red) with respect to centromere. f FISH signals derived from MtR2 (green) and MtR3 (red); note that the MtR3 signal is located between two MtR2 signals on each bivalent. g A merged image of the bivalent chromosomes and the FISH signals. The MtR3 signals coincide with primary constrictions of all chromosomes and also clearly extend to pericentromeric heterochromatin in chromosomes 2, 3 and 5. h,i A pollen mother cell at metaphase I. i Merged image of chromosomes and MtR3 FISH signals. The signals are located on the stretched chromosomal regions. Images of DAPI-stained chromosomes are pseudocolored blue or inverted and in black and white to represent green FISH signals more clearly. Bar represents 5 μm

Full size image

Occurrence of MtR repeats in Medicago R108-1

Although Medicago accession R108-1 has a markedly smaller genome than Jemalong A17, its total chromosome length at pachytene is hardly different: 406 μm for A17 (Kulikova et al. 2001) and 400 μm for R108-1 (data of ten fully analyzed pachytene cells). However, the relative length of the heterochromatic regions is much smaller in R108-1, namely 14.6% in A17 and 8.4% in R108-1. While R108 has only two loci of 5S rDNA (Fig. 4a), Jemalong A17 has three 5S rDNA loci (Gerbach et al. 1999; Kulikova et al. 2001). To determine whether the three MtR repeats occur in the R108-1 genome we used Southern blot analysis. We demonstrated that MtR1 and MtR2 repeats are both absent from the genome of R108-1. However, Southern blot hybridization with MtR3 reveals the characteristic ladder pattern, indicating that this repeat is present in the R108-1 genome as long tandem arrays (Fig. 2b). Furthermore, FISH on meiotic metaphase I chromosomes showed that MtR3 is located at the centromere positions of all chromosomes (Fig. 4b,c). These data are consistent with an essential function of MtR3 in Medicago centromeres.

Fig. 4
figure 4

Fluorescent in situ hybridization signals of repetitive DNA sequences on meiotic chromosomes of R108-1. a Localization of 5S rDNA (red signals) on pachytene bivalents. b A pollen mother cell at late metaphase I. c MtR3 (green signal) is located at the most poleward positions of metaphase I chromosomes. Bar represents 5 μm

Full size image

Quantification of MtR3 regions in different chromosomes

To identify all chromosomes we used the 5S rDNA and MtR1 repeats as diagnostic FISH markers (Kulikova et al. 2001). Seven pachytene complements in which all eight bivalents could be unambiguously identified were selected for quantification. The intensities of each MtR3 signal are expressed as a percentage of the sum of the intensities of all MtR3 regions (Table 1). Given the estimate that MtR3 sequences encompass about 1% of the Medicago genome (560 Mb), they should account for 5.6 Mb. The largest MtR3 region occurs on chromosome 5, accounting for 22.3% of the total signal, or an estimated ∼1.2 Mb of contiguous sequence. The MtR3 region of chromosome 5 is detected not only in the primary constriction, but also clearly defines a heterochromatic block on the short arm (Figs. 3e, 5). The centromere of chromosome 3 is predicted to contain ∼0.9 Mb of contiguous MtR3 sequence. Chromosome 2 has ∼0.8 Mb of MtR3 sequence. For chromosomes 2 and 3 the MtR3 signals were localized on both primary constrictions and pericentromeric heterochromatin blocks. The smallest MtR3 blocks were detected in the centromeres of chromosomes 4 and 7, corresponding to an estimated ∼450 kb (∼8%) each. The centromeres of the other chromosomes (1, 6 and 8) have MtR3 regions in the range of 500–600 kb.

Fig. 5
figure 5

Schematic representation of all chromosomes of Medicago Jemalong A17 chromosomes stretched to equal length and showing the positions of the repetitive sequences MtR1, MtR2, MtR3, and 45S and 5S rDNAs in the (peri)centromeric regions

Full size image
Table 1 Distribution of the MtR3 satellite among centromeres in Jemalong A17 chromosomes
Full size table

Discussion

In this study we used a combination of sequence analysis and FISH to identify satellite repeats located at the centromeres of Medicago chromosomes. The analysis of 20 Mb of whole genome shotgun sequence resulted in the identification of three so-called satellite repeats. As in other eukaryotes these satellite repeats form long arrays with a typical tail-to-head orientation. One of the repeats, MtR3, co-localizes with all primary constrictions. We speculate that MtR3 forms part of the functional centromere domains because it is located at the most poleward position of metaphase chromosomes during meiosis I. The conservation of MtR3 at the centromere of all eight chromosomes in Medicago genotype R108-1 is consistent with this conclusion.

MtR3 occupies regions of different size on individual Medicago chromosomes, varying from ∼450 bp to more than 1.0 Mb. Study of a Drosophila minichromosome has shown that a 420 kb region of centromeric repetitive DNA is the minimal length required for full centromere activity, since shorter regions lead to a progressive reduction of transmission (Sun et al. 1997). A similar minichromosome-based study in maize also concluded that 500 kb of centromeric satellite repeats is the minimum size necessary for fully functional B centromeres (Kaszas and Birchler 1998). Recently the first entire centromeric region of higher eukaryotic organisms has been sequenced in rice chromosome 8 (Nagaki et al. 2004; Wu et al. 2004). Chromatin immunoprecipitation with anti-CenH3 antibodies revealed that CenH3 is associated with the CentO satellite and the CRR family of retrotransposons and the cenH3-containing region spans ∼750 kb (Nagaki et al. 2004). Thus, the region occupied by MtR3 in Medicago is likely to be of sufficient size to function as a centromere.

Previously we calculated that the majority of the genome (80%) is located in the heterochromatic regions (Kulikova et al. 2001). This calculation was based on the length of pachytene chromosomes (406 μm), euchromatic regions (350 μm) and the condensation degree within one euchromatic region (300 kb/μm). Thus the total euchromatic fraction of Medicago DNA was estimated at 105 Mb (300 kb/μm × 350 μm). Using knowledge of genome size for two Medicago accessions and differences in heterochromatin length we can re-evaluate the size of the heterochromatic fraction of the Jemalong A17 genome. If we assume that the condensation degree of euchromatin is similar in both accessions, then 6% difference in heterochromatin length (14.6% of total length of pachytene bivalents in Jemalong A17 and 8.4% in R108-1) corresponds to the difference of 130 Mb between the Jemalong A17 and R108-1 genomes. Thus, heterochromatin accounts for ∼320 Mb of the Jemalong A17 genome, or 60% of the total genome size. This is consistent with the idea that the major part of the Medicago genome is composed of repeats within heterochromatic regions.

We identified three satellite repeats that together comprise 6.5% of genome. The other known tandem repeats in the Medicago genome are 5S rDNA and 45S rDNA. There are three loci of 5S rDNA and one locus of 45S rDNA on Medicago chromosomes (Gerbach et al. 1999; Kulikova et al. 2001). The frequencies of 5S rDNA and 45S rDNA in the 20 Mb random shotgun dataset are 0.25% and 1.2%, respectively. The positions of all highly repetitive DNA sequences are indicated in Fig. 5. The tandem repeats identified to date make up only ∼8% of the Medicago genome, suggesting that several other heterochromatic repeats, likely including transposons and retrotransposons, remain to be characterized.