Novel structural organisation of a Mus musculus dba/2 chromosome shows a fixed position for the centromere

The centromeric domain of most mammalian chromosomes contains repeated DNA sequences (Britten and Kohne, 1968). In the mouse, Mus musculus, the major repetitive DNA component is the satellite DNA sequence family first isolated by Kit (1961). As shown by Pardue and Gall (1969) and Jones (1970) this major satellite is present on all M. musculus chromosomes with the exception of the Y, and constitutes around 10% of the M. musculus genome. Pietras et al. (1983) isolated a second repetitive DNA family from M. musculus and called it the minor satellite because in this mouse species it comprised a much smaller proportion (<1%) of the genome.

By in situ hybridisation Wong and Rattner (1988), Joseph et al. (1989) and Broccoli et al. (1990) have shown that the minor satellite DNA sequences within M. musculus appear to be physically very close to the primary constriction (i.e. the centromere) in these mouse chromosomes. In comparison, the major satellite DNA sequences, although localised to the peri-centromeric heterochromatin in M. musculus, occupy a separate non-overlapping domain to that of the minor satellite DNA sequences (Joseph et al., 1989). This domain, together with its associated C-band, can be deleted from a mouse chromosome without affecting centromere function, which argues strongly that major satellite DNA sequences are not directly involved in centromere function (Broccoli et al., 1990).

In man, the alphoid DNA sequence family is present at the centromeres of all human chromosomes (Mitchell et al., 1985) and, like the minor satellite DNA in M. musculus, is associated with the primary constriction. Conventional simple-sequence satellite DNAs are confined to the more distal C-band heterochromatin in human chromosomes (Gosden et al., 1975; Mitchell et al., 1992). On this evidence, Joseph et al. (1989) proposed that the minor satellite DNAs in M. musculus and the alphoid DNAs in man may have similar roles to play in the biology of centromeres. Anti-centromere autoantibodies present in the sera of patients with the CREST syndrome (Moroi et al., 1980) detect a number of centromere-specific protein antigens in man (Earnshaw and Rothfield, 1985; Earnshaw et al., 1987). The finding by Masumoto et al. (1989) that a 17 base pair sequence (the CENP-B box) present within some monomers of the human alphoid DNAs could specifically bind one of the centromere proteins (CENP-B) recognised by CREST sera, raised the level of interest in a possible functional role for these alphoid monomers in centromere structure. A similar role for the minor satellite DNA in M. musculus was suggested when it was shown that a sequence motif closely related to the human CENP-B box motif was also present within some monomers of this family of repeated DNA sequences (Masumoto et al., 1989).

The one current model of a mammalian centromere is a structure consisting of a number of subunits (Zinkowski et al., 1991), with repetitive DNA sequences giving structural support to proteinaceous components that in turn give rise to the kinetochore structures seen using the electron microscope (Rattner, 1991; Pluta et al., 1990; Cooke et al., 1990). Although this model may indicate one functional role for these repetitive DNA families, it remains unclear what are the precise requirements for an active centromere, and whether the mouse minor or the human alphoid DNAs are directly involved in its formation. The relationship between the CENP-B binding motif and active centromere formation remains obscure, partly because of ultrastructural evidence from HeLa cells (Pluta et al., 1990) where it has been shown that the CENP-B chromosomal protein is located outside the kinetochore itself, and partly from the observation that CENP-B can occur at non-functional centromeres in stable dicentrics (Earnshaw et al., 1989).

To understand the role of the minor satellite in centromere function it is crucial to understand its role in specifying the location of the CENP proteins. Here we report a novel organisation of minor satellite DNA sequences on Chromosome 1 in the inbred mouse strain DBA/2. This chromosome contains two unusually large, well separated blocks of minor satellite DNAs, both containing potential CENP-B binding sites. However, the functional centromere (as defined by CREST staining) is always associated with a sub-region of one block, that closest to the end of the chromosome. This observation suggests that even within a single amplified block the presence of CENP-B binding sites is not sufficient in itself to specify the location of CENP-B. The mechanism by which this might happen is discussed.

DBA/2 and C57BL/6 mice were maintained in Edinburgh. Mice were killed by cervical dislocation and the spleens removed under sterile conditions. Cells were removed from the spleens by forcing 2× 5 mls of RPMI medium through the spleen using a 25 gauge needle. The cells were pelleted at 135 g for 10 minutes and resuspended in 2 ml of RPMI medium. Then 3 ml of phosphate buffered saline (PBS) was added and the suspension carefully loaded onto 3 ml of lymphopaque (Nyegaard Diagnostica, Norway). The sample was centrifuged at 911 g for 20 minutes using a swingout rotor (no brake). The lymphocyte layer was carefully removed and washed in RPMI containing 15% foetal calf serum (FCS) and glutamine (final concentration 0.03%). After pelleting again at low speed the cells were suspended in 10 ml of the RPMI medium containing the above supplements, and 10 ml cultures were set up at a concentration of 10 cells/ml over a 72 hour period at 37°C. Each culture was supplemented with lipopolysaccharide (Sigma) to 0.25mg/ml.

Cells were accumulated in mitosis and metaphase preparations for immunofluorescence work were prepared according to Jeppesen et al. (1992) and references therein. Cells were cultured in the presence of 5-azacytidine as described by Joseph et al. (1989). Chromosomes were fixed in 3:1 methanol:acetic acid and were prepared by standard procedures.

The technique of Koch et al. (1989) in combination with CREST anti-centromere staining (Mitchell et al., 1992) was used with native chromosomes with some minor modifications. The CREST serum used in these experiments was diluted 1/100 in KCM + 10% normal goat serum (NGS) and the second antibody used was a 1/20 dilution of FITC-conjugated goat anti-human antibody + 10% NGS. Denaturation of the chromosomal DNA was carried out using a 30 mM NaOH/1 M NaCl solution at 4°C for 45 minutes. After this time the slide was washed briefly in 10 mM Tris- HCl, pH 7.6, and air-dried. The oligonucleotides used were prepared using an Applied Biosystems DNA synthesiser Model 381A. The primers, numbers 203:

5’-GGAAAATGATAAAAACCACACTGTACAACATATTA-3’ (Pietras et al., 1983), and 204:

5’-CACTTTAGGACGTGAAATATGGCGAGGAAAACTGA-3’ (Horz and Altenburger, 1981), hybridise to the mouse minor and major satellite DNAs, respectively. They show little homology to each other and no detectable cross-hybridisation between these two repetitive DNA families occurs using these oligonucleotides. Oligo C86:

5’-ATTCGTTGGAAACGGGA-3’

corresponds to the mouse CENP-B box (Masumoto et al., 1989).

Digoxigenin-11-dUTP (Boehringer Mannheim) was incorporated during the PRINS reaction and detected with either antidigoxigenin-fluorescein or rhodamine, FAB fragments (Boehringer Mannheim). The annealing temperature was 50°C for 5-10 minutes and extension was carried out at 63°C for 15 minutes. Double PRINS reactions were carried out as described above with the following additional steps. After the first extension at 63°C the slide was immersed in 50 mM EDTA/10 mM Tris-HCl, pH 7.0, for 2 minutes followed by a brief rinse in 10 mM Tris- HCl, pH 7.6. Fresh buffer mix plus the second oligonucleotide was applied but this time with bio-11-dUTP as the reporter molecule in the PRINS reaction. The bio-11-dUTP was detected using a Texas Red-conjugated avidin complex (Vector Labs; see Mitchell et al., 1992). Chromosomes fixed in 3:1 methanol:acetic acid were counter-stained with DAPI (Mitchell et al., 1992). The confocal microscope was used to capture the images of unstained native chromosomes.

Wright’s stain (Wright, 1906) was used to identify the polymorphic chromosome pair prior to carrying out a PRINS reaction with oligo 203. The best banding results were obtained by staining for 2 minutes. The stain was removed during the denaturation step in the PRINS reaction.

Fig. 1A shows the in situ labelling pattern using the minor satellite oligonucleotide 203 as the PRINS reaction primer. Fig. 1B is a different exposure of the same metaphase to show labelling on all the chromosomes. Fig. 1C shows the same metaphase stained with DAPI. One autosomal pair (arrowed in Fig. 1) shows the presence of two large blocks of minor satellite DNA sequences. These blocks appear to be clearly separated by DNA that does not belong to the minor satellite family. The other chromosomes in this metaphase spread give a signal more typical of the minor satellite DNA family in M. musculus strains: paired dot signals close to the centromere/proximal telomere. In this unusual chromosome the more internal minor satellite domain appears to be separated laterally into two blocks whereas the more terminal domain remains more as a single block (see Fig. 1A and B). Oligonucleotide C86 (mouse CENP-B box) gave identical results (not shown).

In order to establish if the minor satellite signals were separated by non-minor satellite DNA sequences, cells were grown in the presence of 5-azacytidine in the culture. As shown in Fig. 2A using oligonucleotide C86 (CENP-B box), the autosomal pair (arrowed) clearly show two minor satellite DNA sequence arrays separated by chromatin, which becomes decondensed by 5-azacytidine, a property associated with mouse major satellite DNA (Joseph et al., 1989). Again, this figure shows that the more internal minor satellite DNA domain is present in two laterally separated blocks, in agreement with the result of Fig. 1A. Fig. 2B shows the same metaphase stained with DAPI.

Double PRINS in situ hybridization carried out on 5-aza- cytidine-treated cells with oligos 203 and 204 (minor and major satellite DNA primers, respectively) is shown in Fig. 3. With oligo 203 the two domains of minor satellite DNA sequences are shown (small arrows in Fig. 3A). Oligo 204 (which recognises the major satellite DNA) gave a signal occupying the remaining area of the centromeric domain (large arrow in Fig. 3B). Fig. 3C is taken with a FITC/Rho- damine dual pass filter (Omega), which superimposes both images and clearly shows that the two domains of minor satellite DNA are separated by major satellite. Fig. 3D is the metaphase stained with DAPI.

In order to identify the autosomal pair containing this minor satellite DNA polymorphism chromosomes were banded using Wright’s stain followed by a PRINS reaction using oligonucleotide 203 (minor satellite): the results are shown in Fig. 4A and 4B. Although the PRINS signal after this treatment is of a lower intensity, Chromosome 1 was identified as the autosome with the two domains of minor satellite DNA sequences (arrowed in Fig. 4A and B).

To determine the position of the centromere on Chromosome 1 PRINS labelling was carried out using oligo C86 (CENP-B box) in combination with indirect immunofluorescence using a CREST anti-centromere serum. The result is shown in Fig. 5. In this confocal image the PRINS signal is simulated red and the immunofluorescence is white (arrowed in Fig. 5). The outline of the unstained native chromosome can be seen. In twenty-five metaphases examined, anti-centromere immunofluorescent labelling was always associated with the more terminal minor satellite DNA block of Chromosome 1. In fact it would appear from these experiments that the functional centromere on Chromosome 1 (as defined by CREST anti-centromere staining) seemed to form at the boundary between this minor satellite block and the adjacent major satellite domain. No examples could be found where CREST anti-centromere labelling was more centrally located within the minor satellite domain.

Polymorphic variants are known to occur within the chromosomes of strains of laboratory mice and in different species of wild mice (Forejt, 1973). Dev et al. (1973) looked at C-band variants within M. musculus strains, and differences in the secondary constrictions of chromosomes in these species were also studied using quinacrine fluorescence (Dev et al., 1971). Both Dev et al. (1973) and Davidson (1989) refer to a C-band variant on Chromosome 1 although no reference was made to the type of DNA sequences involved. These variants appear to be stable with the F1 hybrids inheriting the polymorphic chromosome (Dev et al., 1973). DBA/2 is one of the oldest strains (Forejt, 1973) and a study of its genealogical relationship to the other inbred mice strains (Atchley and Fitch, 1991) shows it to be genetically quite far removed from strains such as C57BL/6. Thus, it is possible that the minor satellite DNA sequence polymorphism found on Chromosome 1 in DBA/2 might have been present in the initial breeding colony. Similar experiments with the mouse strain C57BL/6 and the mouse L929 cell line (data not shown) gave uniform signals on all chromosomes (i.e. a pair of dots at the centromere) when both minor satellite and CENP-B box oligonucleotides were used as PRINS primers. In comparison with the signal from the other autosomes (see Fig. 1A and B) and the signals on both C57BL/6 and mouse L- cell chromosomes (not shown) the increased signal from Chromosome 1 in DBA/2 indicates that amplification of the minor satellite DNA sequences on this chromosome has taken place. Interstrain array size variations prevent the detection of additional unique minor satellite DNA bands on pulse field gels of DBA/2 DNA and C57BL/6 DNA probed with minor satellite oligonucleotides (D. Kipling et al., unpublished).

The chromatin associated with mouse minor satellite DNA has different properties from the chromatin associated with mouse major satellite DNA. This is clearly seen when cells are grown in the presence of 5-azacytidine (Fig. 2A). Under these conditions the chromatin-containing major satellite DNA sequences becomes decondensed. This is in contrast to the chromatin-containing minor satellite DNA sequences, which remain condensed. A similar observation was made by Mitchell et al. (1992) with cultured human lymphocytes. They found that chromatin containing the alphoid DNA sequences remained condensed (as did the CREST antibody binding site) when lymphocytes were grown in the presence of 5-azacytidine. Chromatin containing the simple satellite DNA sequences on the other hand became decondensed. The functional significance of these observations has yet to be determined.

The first proteins specifically to be associated with mammalian centromeres were recognised by autoimmune sera from scleroderma patients characterised by the CREST syndrome (Earnshaw and Rothfield, 1985). Auto-antibodies in CREST patient sera recognise a 80 kDa protein (CENP-B) in both man and mouse (Sullivan and Glass, 1991) as does the CREST serum used in these experiments (not shown). Masumoto et al. (1989) were the first to demonstrate a direct relationship between CENP-B and DNA sequences present within the centromeric domain. They showed that the human alphoid DNAs contained a conserved 17 base pair sequence (the CENP-B box), which formed a specific DNA- protein complex with the CENP-B protein recognised by CREST sera. A closely related sequence was present within the mouse minor satellite DNA. Joseph et al. (1989) and Wong and Rattner (1988) had previously shown that mouse minor satellite DNA was present at the centromeres of all M. musculus chromosomes except the Y. The evidence for this conserved property of human and mouse chromosomes has led to the widespread view that alphoid DNAs of man and the minor satellite DNA of M. musculus have similar functional roles in centromere formation. It is clear from the present study that the centromere, defined by the binding of CREST anti-centromere serum (Cooke et al., 1990), is formed at a specific location on DBA/2 Chromosome 1. Chromosome 1 of this M. musculus inbred strain is atypical, and this naturally occurring variation has enabled us to investigate the relationship between centromeric DNA sequences and centromere biology. The minor satellite arrays on this chromosome are unusual in that they are much larger than the corresponding CREST signal, despite carrying CENP-B box motifs throughout.

Similarly, the more internal array does not stain with CREST despite again containing CENP-B box motifs. Amplification of the minor satellite arrays on Chromosome 1 may have resulted in the accumulation of base mutations within the internal array, leading to sequence changes in the CENP-B box and the negation of a CREST signal. We feel this is unlikely for two reasons. An increase in the stringency of hybridisation of C-86 (CENP-B oligonucleotide) leads to an equal reduction in the hybridisation signal from both the blocks of minor satellite on Chromosome 1 (unpublished observations). If the internal array of minor satellite DNA sequences contained mutated CENP-B binding sites then it would be expected to show much lower labelling than the other one under these conditions. Instead, both blocks behave identically as the stringency of hybridisation is increased. Secondly, studies using variant minor satellite oligonucleotides that anneal to a subset of chromosomes in DBA/2 mice show hybridisation signals of equal intensity to both minor satellite arrays on Chromosome 1 (Kipling et al., unpublished). Both observations strongly suggest sequence homogeneity between the minor satellite arrays on DBA/2 Chromosome 1.

Two well-characterised aspects of centromere function, CREST staining and sister chromatid attachment, are found for one domain and both are absent from the other. We hypothesise that epigenetic control of CENP-B binding causes it to be found only in a subdomain of the more terminal of the two minor satellite DNA arrays, and results in both CREST staining and sister chromatid attachment. This epigenetic control of CENP-B binding might be methylation as the CENP-B box contains two CpG dinucleotides, and minor satellite DNA is known to be heavily methylated in vivo (Chapman et al., 1984). Broccoli et al. (1990) also considered methylation as a possible mechanism for the formation of centromeres. However, the sensitivity using antibodies against 5-methyl cytosine did not allow any definite conclusions to be drawn, mainly because of interference from the large domain of major satellite DNA sequences. Chromosome 13, which had lost almost all its C-band material still showed a small signal with the anti-5-methylcyto- sine antibody suggesting that the minor satellite DNA sequences at this position were methylated (Broccoli et al., 1990).

An alternative explanation is that additional, but as yet undiscovered, DNA sequences might be required for the formation of an active centromere. Vig and Richards (1992) have argued that the formation of the primary constriction in some mouse chromosomes (their type 1 chromosome) does not always require the presence of mouse minor satellite DNA sequences. Their data, however, is based solely on the technique of in situ hybridisation to detect the presence of minor satellite DNA sequences on chromosomes and the absence of a signal may well be caused by the lack of sensitivity of this technique in detecting a few copies of a DNA sequence. However, the fact that the Y chromosome in M. musculus lacks minor satellite sequences yet possesses a functional centromere, together with evidence that the CENP-B protein itself appears to be localised beneath the kinetochore plates (Cooke et al., 1990), suggests that it is possible for other DNA sequences to be more directly associated with the structures that comprise the kinetochore.

Another centromere function is the association of sister chromatids. The more internal block of minor satellite on DBA/2 Chromosome 1 does not appear to be demonstrating this feature, as the blocks on sister chromatids are laterally separated. In contrast the blocks of the more terminal array appear to be very close together on sister chromatids and are often seen as a single fused block. A similar observation has been made by Broccoli et al. (1990) using a mouse L929 marker chromosome. Here an interstitial minor satellite array that was not active as a centromere was always separated on sister chromatids, whereas the minor satellite DNA sequences at the active centromere were more closely associated. Together these observations suggest that the presence of minor satellite DNA may not in itself be sufficient to confer this aspect of centromere function.

We thank Professor H. J. Evans and Drs Peter Jeppesen and Howard Cooke for their useful comments on this manuscript. One of us (D.K.) is a Beit Memorial Fellow.