Breakpoint cluster regions of the AML-1 and ETO genes contain MAR elements and are preferentially associated with the nuclear matrix in proliferating HEL cells

Chromosomal rearrangements frequently occur in some specific places (`hot spots') in the genome. These recombination hot spots are usually separated by 20-100 kb regions of DNA that are rarely involved in rearrangements (Henglein et al., 1989; Geng et al., 1993). A correlation between the above distances and the average size of DNA loops fixed at the nuclear matrix seems to be quite likely. Recent studies have demonstrated that DNA loop anchorage regions can be fairly long and can harbor DNA recombination hot spots (Svetlova and Razin, 2001). One of the major protein components of the nuclear matrix is DNA topoisomerase II (Topo II), which displays an intrinsic intermolecular DNA ligation activity and can mediate illegitimate DNA recombination in vitro (Gale and Osheroff, 1992). Exposure of mammalian cells to Topo-II-specific drugs stimulates different genomic rearrangements such as deletions, insertions and translocations (Maraschin et al., 1990; Shibuya et al., 1994). In agreement with this, chemotherapy of tumors with Topo II-specific drugs (such as VP16 or m-AMSA) frequently causes secondary leukemia originating as a result of chromosomal rearrangements (Auxenfants et al., 1992; Super et al., 1993) (for a review, see Rowley, 1993). On the basis of these data, we proposed that illegitimate DNA recombination occurs preferentially at sites of DNA contact with the nuclear matrix (Razin, 1999). Consequently, illegitimate DNA recombination was thought to result predominantly in the loss or repositioning of entire DNA loops. Similar ideas are shared by others (Felix, 1998; Ahuja et al., 2000; Lovett et al., 2001; Whitmarsh et al., 2003). The data of the above-cited authors clearly show the role of Topo II in chromosomal rearrangements. It is, however, less clear what determines the positions of recombination hot spots. According to our view, the location of DNA in respect to the nuclear matrix should be of primary importance. Correspondingly, we argued that high-salt-insoluble Topo II of the nuclear matrix is the main target for a range of Topo-II-specific anticancer drugs (Gromova et al., 1995a; Iarovaia et al., 1996) and that it is the inhibition of this enzyme that stimulates chromosomal rearrangements (Razin, 1999). In order to verify our point of view, we here report a study of the association with the nuclear matrix of two breakpoint cluster regions present in the AML-1 and ETO genes that are frequently involved in reciprocal recombinations (Elsasser et al., 2003; Zhang et al., 1994; Le et al., 1998; Davis et al., 2003; Westendorf et al., 1998). Both genes contain relatively short breakpoint regions where Topo-II cleavage sites and DNase-I-hypersensitive sites have been mapped (Zhang et al., 2002; Bystritskiy and Razin, 2004). In the present study, we investigate whether the AML-1 and ETO breakpoint cluster regions are located at the nuclear matrix in cells where these genes are not rearranged. Translocations t(8;21) resulting in the fusion of AML-1 and ETO genes are among the most frequent chromosomal rearrangements observed in acute myeloid leukaemia (Davis et al., 2003).

Human erythroleukemia cells, HEL 92.1.7 (ATCC), were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum.

Cells were pelleted (700 g, 5 minutes), washed twice with RPMI 1640 medium and resuspended in permeabilization buffer [10 mM PIPES (pH 7.8), 100 mM NaCl, 3 mM MgCl₂, 0.5 mM phenylmethylsulfonyl fluoride (PMSF), 0.1 mM CuSO₄, 300 mM sucrose and 0.5% (v/v) Triton X-100] to a final concentration of 2×10⁶ cells ml^–1. After 4 minutes of incubation on ice, the cells were pelleted onto silane-coated microscope slides using a Cytospin centrifuge. In some experiments, the permeabilized cells were treated with RNase A [25 μg ml^–1 in 10 mM PIPES (pH 6.8), 10 mM EDTA, 0.05 mM spermine, 0.125 mM spermidine, 0.1% (w/v) digitonin]. The cells on the slides were then treated (4 minutes at 0°C) with high-salt solution [2 M NaCl, 10 mM PIPES (pH 6.8), 10 mM EDTA, 0.05 mM spermine, 0.125 mM spermidine, 0.1% (w/v) digitonin]. After this treatment, the slides were sequentially washed (1 minute for each wash) with 10×, 5×, 2× and 1× PBS, and then with 10%, 30%, 70% and 96% ethanol. The air-dried slides were fixed in methanol/acetic-acid (3:1) mixture and baked at 70°C for 1 hour.

Nuclear halos were treated sequentially with RNase A (100 μg ml^–1 in 2× SSC) and pepsin (0.01% in 10 mM HCl), post-fixed with 1% paraformaldehyde in 1× PBS and rinsed sequentially in 70%, 80% and 96% ethanol. To denature DNA, the slides were incubated in 70% formamide, 2× SSC solution for 5 minutes at 74°C, dehydrated in cold 70%, 80% and 96% ethanol, and air dried.

Hybridization probes were labeled with biotin-16-dUTP using a `Biotin high-prime' kit (Roche). The hybridization mixture contained (in a final volume of 10 μl) 50% (v/v) formamide, 2× SSC, 10% dextran sulfate, 0.1% Tween-20, 10 μg sonicated salmon-sperm DNA, 10 μg yeast tRNA and 25-50 ng of a labeled probe. Before hybridization, the mixture was incubated for 10 minutes at 74°C to denature DNA. Hybridization was carried out overnight at 40-45°C. After hybridization, the samples were washed twice in 50% formamide, 2× SSC at 43-48°C for 20 minutes.

The biotinylated probe was visualized using anti-biotin monoclonal antibodies conjugated with Alexa 488 (Molecular Probes) with subsequent signal amplification using an Alexa-488 signal-amplification kit for mouse antibodies (Molecular Probes). In some experiments, two additional layers of antibodies (chicken anti-goat and goat anti-chicken), both conjugated with Alexa-488, were used. In all cases, the DNA was counterstained with DAPI (4′,6-diamidino-2-phenylindole). The results were examined under a fluorescence Axioplan microscope (Opton) and recorded using a cooled charge-coupled-device AT200 camera (Photometrics, Tucson, Arizona).

To isolate nuclear matrices, the cells were washed with cold TM buffer [10 mM Tris-HCl (pH 7.4), 3 mM MgCl₂, 1 mM PMSF] supplemented with 0.2 mM CuSO₄ and resuspended in the same buffer. Then, 5% Nonidet P-40 was added up to a final concentration of 0.1% and the suspension was incubated on ice for 10 minutes. This was followed by two washes with TM buffer. Permeabilized nuclei were then resuspended again in TM buffer and DNase I was added up to 100 μg ml^–1. After incubation for 30 minutes at 37°C, an equal volume of ice-cold extraction buffer [4 M NaCl, 20 mM EDTA, 20 mM Tris-HCl (pH 7.4)] was added. After incubation for 20 minutes at 0°C, the nuclear matrices were precipitated by centrifugation for 15 minites at 1000 g and 4°C. The pellet was washed once with 0.5 × extraction buffer and twice with TM buffer supplemented with 0.25 mM sucrose. The matrices were stored at –20°C in TM buffer supplemented with 0.25 mM sucrose and 50% glycerol. The matrix-attachment region (MAR) assay was carried out exactly as described by Cockerill and Garrard (Cockerill and Garrard, 1986). The matrix-bound DNA was purified by a conventional procedure and analysed using electrophoresis in 1% or 1.5% agarose gels. Digestion of cloned DNA by restriction enzymes and labeling of the DNA fragments were carried out as described previously (Maniatis, 1982).

Total RNA (1 μg) treated with DNase I [polymerase-chain-reaction (PCR) grade] (Gibco/Life Technologies) was reverse transcribed into cDNA [using non-specific short primers and M-MuLV reverse transcriptase (MBI)]. The test fragments of the AML-1 and ETO genes were PCR amplified with Taq DNA polymerase using the above cDNA as a template and the following primers: AML-1, TGAGGGTTAAAGGCAGTGGA and AGATGATCAGACCAAGCCCG (product length, 156 bp); ETO, AGTGCAACTGGGTCTGGGTT and CTGCATAATGGACATGGTAG (product length, 192 bp). As a positive control, the same primers were used to amplify the corresponding test fragments on a total genomic DNA template.

A modification of the previously described experimental protocol (Maya-Mendoza and Aranda-Anzaldo, 2003) was used. Briefly, cells prelabeled with ³H-thymidine were lysed in buffer A [10 mM Pipes pH (7.8), 100 mM NaCl, 3 mM MgCl₂, 0.1 mM CuSO₄, 0.5 mM PMSF, 300 mM sucrose, 0.5% Triton X-100] for 20 minutes at 4°C. After two washes in buffer B [50 mM Tris-HCl (pH 8.0), 3 mM CaCl₂], the nuclei were resuspended in 1 ml of the same buffer (10⁶ nuclei ml^–1) and then treated with micrococcal nuclease (Fermentas; 7.5 units ml^–1). Digestion was carried out at 37°C for 30 minutes. After digestion, a 2 ml volume of cold 1.5× extraction buffer (3 M NaCl, 30 mM EDTA) was added and the mixture was incubated for 20 minutes at 4°C. The nuclear matrices were precipitated and washed twice with cold extraction buffer. Then, matrix-bound DNA was isolated and the size distribution of the DNA fragments was analysed by agarose-gel electrophoresis. Total nuclear DNA (used as a control template for PCR amplifications) was digested to fragments with the same average size as matrix DNA fragments. The same amounts of total and nuclear matrix DNA were used as templates in parallel PCR amplifications of the test fragments. The primers used to carry out PCR amplifications are shown in Table 1. The amplified DNA fragments were separated by electrophoresis and the gels were scanned to estimate the relative quantity of DNA in each band. In preliminary experiments, the optimal PCR conditions permitting an accurate estimation of the differences in the quantities of template DNA over a 1× to 10× range were determined (usually 20 PCR cycles).

Visualization of specific DNA sequences on nuclear halos by hybridization opens the possibility of studying their partitioning between matrix-bound and looped-out DNA fractions (Balajee et al., 1996; Gerdes et al., 1994; Ratsch et al., 2002). In our recent experiments, this method was successfully used to study the partitioning of the human dystrophin gene into loop domains (Iarovaia et al., 2004). Importantly, a good correlation was observed between results of biochemical mapping experiments and in situ hybridization experiments. Thus, DNA fragments mapped as loops by Topo-II-mediated DNA loop excision (Razin et al., 1993; Razin et al., 1991) were indeed looped out on nuclear halos, and DNA fragments mapped by the above protocol as DNA-loop anchorage regions were present almost exclusively on the nuclear matrix when hybridized to nuclear halos (Iarovaia et al., 2004).

In the present study, hybridization in situ with nuclear halos was used to characterize the spatial positions of AML-1 and ETO BCRs. There are several BCRs in AML-1 involved in reciprocal recombinations with different partners (Zhang et al., 2002). All BCRs participating in recombinations with ETO-1 are located in intron 5 (Zhang et al., 2002), in which three subclusters can be recognized. For our studies, we have selected BCR3, which is the closest to exon 6. We PCR amplified a 3009 bp DNA fragment mapped to position 88172-91180 of the genomic AML-1/RUNX-1 sequence (GenBank accession no. AP001721). This fragment (here termed AML-BCR3), which does not contain any repetitive sequences, was used as a probe for fluorescent in situ hybridization (FISH) with nuclear halos (prepared from HEL cells as described in Materials and Methods).

It should be mentioned that, although HEL cells originate from a human erythroleukaemia, they do not bear t(8;21) translocations that affect AML-1 or ETO (Erickson et al., 1996). The results of hybridization are shown in Fig. 1A and Table 2. One can see that the AML-BCR3 probe hybridized preferentially, but not exclusively, within the nuclear matrix region. 77% of identified signals were present on the nuclear matrix and 23% on the DNA-loop halo. The borders of the nuclear matrix were determined by immunostaining of lamins A and B (Fig. 2A). The much more intensive staining by DAPI of the nuclear matrix compared with the crown of DNA loops does not reflect the real DNA distribution between the nuclear matrix and the loop halo (perhaps because of a non-specific adsorption of the stain on nuclear proteins). To characterize more accurately the distribution of DNA in nuclear halos, we have carried out hybridization in situ with an abundant interspersed repetitive sequence (the Alu repeat). The results of this experiment (Fig. 2B) show that hybridization signals are not preferentially concentrated within the nuclear matrix area. To a first approximation, the distribution of hybridization signals between the nuclear matrix and the crown of DNA loops was random [i.e. the proportions of Alu repeats present in the nuclear matrix and in the loop halo roughly reflected their areas (25-30% and 70-75%, correspondingly)]. Hence, the observed proportion of the AML-BCR3 signals on the nuclear matrix is ∼2.5 times higher than expected in case of a random distribution of signals between the nuclear matrix and the loop halo. Interestingly, a significant proportion of the observed signal was close to the nuclear periphery (Table 2).

In the ETO gene, three BCRs were found between exons 1a and 1b (Zhang et al., 2002). We have chosen for further studies the BCR2 located in the middle of intron 1b. A corresponding DNA fragment (here termed ETO-BCR2) 2.75 kb in length with the coordinates 108459-111173 on the NT_034899 sequence (region 136363-251951) was PCR amplified and cloned. After labeling of the insertion with biotin, in situ hybridization was carried out with nuclear halos from HEL cells (Fig. 1B, Table 1). It is evident that ETO-BCR2 is localized for the most part to the nuclear matrix.

Importantly, in no case was the distribution of observed hybridization signals affected by pretreatment of nuclear matrices with RNase A (Fig. 1A′,B′). Thus, the observed localization of BCRs on the nuclear matrix was not due to artificial coprecipitation with nascent transcripts (see below).

Previous studies have demonstrated that transcribed genes interact transiently with the nuclear matrix. Thus, it was important to know whether the ETO and AML-1 genes are transcribed in the HEL cells used in our experiments. In order to answer the question, the reverse-transcription PCR (RT-PCR) approach was used. A reverse-transcription reaction was carried out on total cellular RNA using random primers and then the test fragments from the ETO and AML-1 genes were PCR amplified. In both cases, clear bands of the expected sizes were observed (Fig. 3). Hence, both genes are transcribed in HEL cells.

One trivial explanation of the results of in situ hybridization with nuclear halos (Figs 1, 2) is that both genes, including BCRs, are attached to the nuclear matrix solely because they are transcribed. Indeed, the association of transcribed DNA sequences was reported by many researchers (for a review, see Razin, 1987). In order to find out whether the whole AML-1 and ETO genes (i.e. not only their BCRs) are attached to the nuclear matrix, unique regions located at a distance of about 65 kb from BCRs were selected in both genes – a DNA fragment 3.46 kb in length (here termed ETO-control) with the coordinates 26553-29991 on the NT_034899 sequence (region 136363-251951) and a DNA fragment 1.6 kb in length (here termed AML-control) with coordinates 117400-119009 on the genomic AML-1/RUNX-1 sequence (GenBank accession no. AP001721). These DNA fragments were PCR amplified and cloned. Hybridization of these two probes to nuclear halos demonstrated their preferential localization to the crowns of DNA loops (Fig. 4A,B, Table 2). In both cases about 75% of signals were detected in loops, a figure similar to that observed previously when probes from dystrophin loops were hybridized to nuclear halos (Iarovaia et al., 2004).

Although hybridization in situ with nuclear halos of probes located far from BCRs but within transcribed regions of both genes did not favor the above possibility, an additional experiment was done in order to clarify the situation. We used a biochemical PCR-based approach (Maya-Mendoza and Aranda-Anzaldo, 2003) to study the relative representation of different regions of the AML-1 gene (including BCR3 and distant transcribed parts of this gene) in total DNA and in nuclear matrix DNA. The AML-1 gene was chosen for this analysis because, according to the RT-PCR analysis (see above; Fig. 5) in HEL cells, it is transcribed much more intensively than the ETO gene. Nuclear matrices were obtained by a modification of the sequential extraction procedure (Berezney and Coffey, 1977) using limited treatment of nuclei with staphylococcal nuclease (Razin et al., 1979) The average size of the nuclear matrix DNA fragments was about 1 kb, and about 2% of total DNA was recovered in the nuclear matrix. The same amounts of total DNA and nuclear matrix DNA were used for PCR amplification of the test fragments scattered along the AML-1 gene (Fig. 5). The results of amplifications are shown in Fig. 5 below the scheme. It is evident that only the test fragment derived from BCR3 is enriched in nuclear matrix DNA (at least seven times enrichment compared with total DNA, according to the results of scanning of the bands). Other test fragments were distributed almost equally in total DNA and matrix DNA. This result clearly demonstrates that the attachment of this BCR to the nuclear matrix is not a result of AML-1 gene transcription, as the other studied regions are equally transcribed but are not over-represented in nuclear matrix DNA.

MARs are eukaryotic DNA sequences that bind in a specific fashion to the nuclear matrix in the presence of a vast excess of competitor prokaryotic DNA (Cockerill and Garrard, 1986). MARs are likely to participate in DNA-loop anchorage to the nuclear matrix, although most of them represent potential attachment sites (Iarovaia et al., 1996; Razin, 2001). In order to find out whether the BCRs studied here contain MARs, a standard in vitro binding assay (Cockerill and Garrard, 1986) was used. The bona fide MAR from the Drosophila histone gene cluster (Cockerill and Garrard, 1986; Mirkovitch et al., 1984) was used as a positive control. The results of the experiments are presented in Fig. 6. The 2.7 kb ETO-BCR2 fragment was cut into two subfragments of 1.8 kb and 0.9 kb. The largest of these two fragments (1.8 kb) was not bound by the nuclear matrix (the binding was competed by the same amount of competitor DNA as the binding of the linearized pUC vector). By contrast, the 0.9 kb fragment was bound by the nuclear matrix to the same extent as the bona fide MAR from the Drosophila histone gene domain. Hence, this fragment contained a strong MAR. The 3 kb AML-BCR3 fragment was cut into 1.8 kb and 1.2 kb subfragments (Fig. 6). Again, these fragments were mixed with 2.8 kb pUC18 DNA (negative control) and a cloned 1.7 kb MAR from the Drosophila histone gene cluster (positive control) and a standard matrix-binding assay was carried out. It is evident (Fig. 3B) that the 1.8 kb fragment has the same affinity to the nuclear matrix as the MAR from the Drosophila histone gene cluster. By contrast, the 1.2 kb fragment was not at all bound by the nuclear matrix. Summarizing, we conclude that both ETO-BCR2 and AML-BCR3 contain MARs.

The mechanisms of illegitimate recombination resulting in chromosomal translocations are not yet fully understood. Even less is known about the reasons for non-random distribution of recombination sites along chromosomes. We proposed that the nuclear architecture might play a certain role in determining the positions of recombination hot spots (Razin, 1999). In particular, it was proposed that these hot spots might be localized within matrix attachment regions (Razin, 1999). This idea was corroborated by some indirect evidence. Namely, it was found that Topo II of the nuclear matrix, which seems to be in contact with DNA in the MARs, is a principal target for different anticancer drugs (Fernandes and Catapano, 1995; Gromova et al., 1995a; Lambert and Fernandes, 2000). Inhibition of Topo II activity induces double-strand breaks in DNA. Mistakes in the course of repair of these breaks might cause illegitimate recombination. It is also worth mentioning that nuclear MARs constitute weak points in chromosomes where DNA cleavage by different agents frequently occurs (Gromova et al., 1995b). Being far from each other on the DNA chain (and even in different chromosomes), the MARs might reside close to each other at the nuclear matrix (i.e. in physical space) and this will enhance the probability of illegitimate recombination between distal chromosomal regions as well as between different chromosomes. In a previous study, we demonstrated that, in the Chinese hamster genome, the recombination hot spot located close to the GNA3 gene resided within the matrix attachment area (Svetlova et al., 2001). Here, we have demonstrated that the BCRs of the AML-1 and ETO genes are preferentially (and, in the case of the ETO gene, almost exclusively) located at the nuclear matrix. It is important that the association of these BCRs with the nuclear matrix is not a consequence of transcription. First, probes located far from BCRs but within the transcribed parts of both genes hybridized preferentially to the loop halos (in contrast to BCR probes, which hybridized preferentially within the nuclear cores). Second, using a semiquantitative PCR-based approach (Maya-Mendoza and Aranda-Anzaldo, 2003), we have demonstrated that, in contrast to BCRs, other transcribed regions of the AML-1 gene are not enriched in nuclear matrix DNA. These results are in apparent contradiction to some earlier observations showing that transcribed DNA sequences are over-represented in nuclear matrix DNA (Ciejek et al., 1983; Robinson et al., 1983). However, in these previous studies, hybridization methods were used that integrated the signal from long DNA fragments. Furthermore, the loop DNA was cut from the nuclear matrix by restriction nucleases, and thus the nuclear matrix DNA was composed of relatively long DNA fragments. If, indeed, transcribed DNA sequences became temporarily associated with the nuclear matrix via matrix-bound transcription complexes (Jackson and Cook, 1985; Razin, 1987), the intensity of a signal observed upon hybridization of a corresponding DNA probe with the nuclear matrix DNA would depend on the length of the nuclear matrix DNA fragments and on the intensity of transcription. Thus, long DNA fragments with several elongating RNA polymerase II complexes would behave as though they were bound to the nuclear matrix. By contrast, the PCR-based approach used in the present study permits the spatial positions of short DNA fragments to be analysed. 300 bp test fragments would be over-represented in nuclear matrix DNA only if they were attached to the nuclear matrix in most cells in the population. Thus, the association of BCRs with the nuclear matrix is likely to be permanent and not transient. In agreement with this conclusion, we have demonstrated that both BCRs studied in this work contain MARs that participate in the anchorage of DNA loops to the nuclear matrix (Iarovaia et al., 1996). The MARs are characterized by the presence of different simple motives and imperfect repeats (Boulikas, 1995) and thus present perfect targets for illegitimate recombination carried out by both non-homologous and homologous end-joining systems. According to recent studies, in untransformed human cells, each interphase chromosome occupies a specific radial position within the nuclear space (Boyle et al., 2001). Spatial proximity of translocation-prone loci might increase the probability of reciprocal recombination between them (Roix et al., 2003; Parada et al., 2002). Thus, chromosomes with similar radial positions in the nuclei are more frequently involved in reciprocal recombinations than are chromosomes with different radial positions. It is interesting that chromosome 8 (harboring the ETO gene) is characterized by a relatively distal radial position in normal nuclei, whereas chromosome 21 (harboring the AML-1 gene) is characterized by one of the most central radial positions (Boyle et al., 2001). Hence, the expected frequency of reciprocal recombinations between these chromosomes should be rather low. The preferred location of a significant portion of AML-BCR3 on the peripheral layer of the nuclear matrix might explain the apparent contradiction. The chromosomal territories are rather amorphous, dynamic structures and the location of a particular gene outside the main body of a territory was observed by several authors (Mahy et al., 2002; Ragoczy et al., 2003). It is natural that, being fixed (even temporarily) at the nuclear periphery, AML-BCR3 has a much better chance to meet the ETO-BCRs than any chromosomal region located close to the center of the nucleus.

We are grateful to R. Hancock for critical reading of the manuscript and helpful discussion. This work was supported by a grant of the MCB program of the Presidium of the Russian Academy of Sciences and by RFFI grants 02-04-48369 and 03-04-48627.