Unité de Cytogénétique Moléculaire et Oncologie (UMR 147 CNRS), Institut Curie, 26 rue d'Ulm, 75248 Paris Cédex 05, France

Correspondence to: M Debatisse, E-mail: michelle.debatisse@curie.fr

Gene amplification is frequently associated with tumor progression, hence, understanding the underlying mechanisms is important. The study of in vitro model systems indicated that different initial mechanisms accumulate amplified copies within the chromosomes (hsr) or on extra-chromosomal elements (dmin). It has long been suggested that formation of dmin could also occur following hsr breakdown. In order to check this hypothesis, we developed an approach based on the properties of the I-SceI meganuclease, which induces targeted DNA double-strand breaks. A clone containing an I-SceI site, integrated by chance close to an endogenous dhfr gene locus, was used to select for methotrexate resistant mutants. We recovered clones in which the I-SceI site was passively co-amplified with the dhfr gene within the same hsr. We show that I-SceI-induced hsr breakdown leads to the formation of dmin and creates different types of chromosomal rearrangements, including inversions. This demonstrates, for the first time, a direct relationship between double-strand breaks and inversions. Finally, we show that activation of fragile sites by aphidicolin or hypoxia in hsr-containing cells also generates dmin and a variety of chromosomal rearrangements. This may constitute a valuable model to study the consequences of breaks induced in hsr of cancer cells in vivo.

Oncogene (2002) 21, 7671-7679. doi:10.1038/sj.onc.1205880

I-SceI meganuclease; mammalian gene amplification; double-strand break repair; fragile sites; hypoxia; chromosomal rearrangements

Introduction

Gene amplification is one of the most frequent alterations found in cancer cells and has been repeatedly correlated with over-expression of many different cellular oncogenes. Because this phenomenon contributes to tumor progression and tumorigenesis (Brison, 1993; Schwab, 1999), it is of importance to identify the mechanisms driving gene amplification in mammalian cells. In cells from advanced tumors, the reconstitution of these mechanisms is obscured by the occurrence of secondary rearrangements (Knudson, 2000; Nowell et al., 1998). To avoid this problem, model systems were developed of cultured cells selected in vitro to resist inhibitors of enzymes essential for cell growth. This has enabled analysis of mutant cells a few generations after the initial event triggering the amplification process (Smith et al., 1990).

These studies led to the identification of at least two different initial gene amplification mechanisms (Coquelle et al., 1997; Carroll et al., 1988; Ma et al., 1993; Ruiz and Wahl, 1988; Smith et al., 1992; Toledo et al., 1992, 1993; Trask and Hamlin, 1989; Windle et al., 1991). The breakage-fusion-bridge (BFB) cycle mechanism, first described by Mcclintock (1942), accumulates copies organized as large inverted repeats on a chromosome arm where one normal gene copy maps in non-amplified cells (Coquelle et al., 1997; Ma et al., 1993; Toledo et al., 1992, 1993). The operation of this mechanism has been associated with the observation of fused sister chromatid, anaphase bridges and deletion of the distal part of the amplified chromosome arm (Figure 1) (Coquelle et al., 1997; Toledo et al., 1993). We also demonstrated that fragile sites, which are regions where recurrent breaks are detected in metaphase chromosomes when cells are grown under stressful conditions (Richards, 2001; Sutherland et al., 1998), determine the size and the sequence content of early amplicons (Coquelle et al., 1997).

Intermediates of BFB cycles, such as anaphase bridges and fused sister chromatids, were observed in cells amplified for different oncogenes (Gisselsson et al., 2000; Saunders et al., 2000), establishing the key role of BFB cycles during tumor progression (Knudson, 2000). More recently, the involvement of fragile sites in oncogene amplification in human cells as also been described (Hellman et al., 2002).

Studies on model systems showing that cells containing dmin can be completely free of chromosomal rearrangements revealed the existence of an independent amplification mechanism, relying on looping out of megabase-long extra-chromosomal DNA sequences (Figure 1) (Coquelle et al., 1997; Toledo et al., 1992, 1993). Other observations correlated the presence of dmin to chromosome breakage (Ruiz and Wahl, 1988; Windle et al., 1991) but whether these features result from the operation of a particular mechanism or from secondary events leading to hsr disintegration is still debated. Thus, the nature of the mechanism(s) generating dmin in vitro and in vivo remains controversial.

In all model systems, rearrangements of the amplified copies were observed upon clonal expansion. For example, dmin can fuse and reintegrate into the chromosomes, giving rise to secondary hsr (Coquelle et al., 1998; Ruiz and Wahl, 1988; Windle et al., 1991), and it was demonstrated that these events are stimulated by DNA double-strand breaks (DSB) (Coquelle et al., 1998). On the other hand, the formation of dmin following breakdown of an hsr has also been proposed (Balaban-Mallenbaum et al., 1981; Singer et al., 2000). The aim of this study was to determine unambiguously whether such an inter-conversion pathway operates efficiently and whether double-strand breaks activate this putative process. For that purpose, we developed an experimental approach allowing us to generate multiple DSBs within an hsr by using a mitochondrial group I intron-encoded nuclease of Saccharomyces cerevisiae (I-SceI) (Dujon, 1989). The 18 bp recognition sequence of I-SceI (IS) is absent from most genomes and can be introduced in mammalian cells upon transfection (Rouet et al., 1994). I-SceI expression in cells containing IS induces site-specific DSBs (Jasin, 2000). Here, we isolated cells in which numerous copies of IS lie within an hsr, and we show for the first time that hsr breakdown resulting from multiple DSBs induced by Isce-I expression gives rise to dmin and/or chromosomal rearrangements. This clearly demonstrates that dmin can form through hsr breakage and points to the role of DSBs in triggering this phenomenon. Among the rearrangements observed, inversions were recurrently observed, which demonstrates that DSBs actively contribute to generate this category of chromosomal rearrangements.

We next studied the consequences of induction of fragile sites in cells containing an hsr. We previously demonstrated that oxygen privation (hypoxia), an important parameter of the tumor microenvironment (Brizel et al., 1996; Graeber et al., 1996; Kinzler and Vogelstein, 1996), which enhances gene amplification frequency (Luk et al., 1990; Rice et al., 1986), induces breaks at fragile sites (Coquelle et al., 1998). Here we show that activation of fragile sites, by drugs such as aphidicolin or by hypoxia, induces hsr breakdown as do I-SceI induced DSBs. Therefore, various stresses are able to promote both the interconversion of hsr into dmin, and the formation of most types of chromosomal rearrangements frequently associated with late stages of gene amplification.

Results

Design of a cellular model permitting induction of site-specific DSBs within an hsr

In order to address the question of the possible generation of dmin from breakdown of a pre-existing hsr, we took advantage of a previously selected derivative of the GMA32 Chinese hamster cell line, named 112. Briefly, GMA32 cells were transfected with a plasmid bearing the I-Sce1 site (IS) and the neo gene. G418 resistant clones were selected for, and we mapped integrated plasmids in a large number of independent clones by using the fluorescent in situ hybridization technique (FISH). In cells of clone 112, a single copy of the plasmid was integrated by chance on the p arm of one chromosome 2, about 3 megabases telomeric to the dhfr locus, without further rearrangement of that chromosome (Pipiras et al., 1998).

We reasoned that because of the localization of IS relative to the dhfr gene, methotrexate (MTX)-induced dhfr amplification events should lead to a passive co-amplification of IS. Thus, cells of line 112 were plated in MTX-containing medium at a concentration allowing the recovery of mutants amplified for the dhfr gene, without expression of the I-sce1 enzyme. Clones were recovered at similar frequencies, regardless of whether the gene was amplified from the chromosome 2 bearing the plasmid or from the normal chromosome 2. Moreover, the localization of the extra-copies of the dhfr gene on the p arm of chromosome 2, and their organization relative to the position of the characteristic DAPI band of this chromosome arm (an example is shown in Figure 2), were similar in the two categories of clones as well as in previously described mutants selected from the parental GMA32 cell line (Pipiras et al., 1998). All these observations indicate that the transgene was neither initiating nor driving the amplification process in these experimental conditions.

We further analysed two clones, 112-MTX1 and 112-MTX2, in which the plasmid containing IS was passively co-amplified with the dhfr gene within an hsr exhibiting the structural features that characterize the operation of BFB cycles (Figure 2). In both mutants, we observed, from centromere to telomere of the amplified chromosome arm, one dhfr copy at its wild type location, a typical duplication of a dark DAPI band specific to a normal 2p arm, and a variable number of extra-copies of the dhfr gene. Figure 2B1 shows an example of such an hsr in 112-MTX1 cells. As expected from the joint operation of the BFB cycles and of superimposed mechanisms leading to rearrangement and shortening of the initial amplicons (Toledo et al., 1992), the relative numbers of dhfr genes and of the IS-containing plasmid within the hsr vary from cell to cell in each clone. We generally detected three to seven copies of the plasmid within the hsr, as compared to tens of extra-copies of the dhfr gene (see Figure 2B for examples in clone 112-MTX1). In all the cells, the extra-copies of the dhfr gene were distributed up to the end of the amplified 2p (Figure 2B1), while significant variations were seen in the localization of the copies of the plasmid along the hsr of different cells (Figure 2B2,3). However, the existence of cells containing several IS within an hsr offered a convenient model to study the consequences of multiple site-specific DSBs occurring on a chromosome arm.

Formation of dmin through DSB-mediated hsr breakdown

Clones 112-MTX1 and 112-MTX2 were transfected with a control plasmid (pcDL-SR296) or with a plasmid coding the I-SceI nuclease (pI-SceI) (Pipiras et al., 1998). Experiments based on cotransfection of either plasmid with a plasmid coding for the green fluorescent protein (gfp) indicated that the transfection efficiencies routinely reached 20-25% (see Material and methods). Unless specifically mentioned, MTX was present in the culture medium at all steps of the experiments in order to select for cells that retain amplified copies of the dhfr gene. Five days after transfection, the cells were recovered and studied by FISH with cosmid probes specific for the dhfr locus. Following transfection with pI-SceI, about 25% of the cells demonstrated marked rearrangements associated with changes in the organization of the hsr. In particular, extrachromosomal copies of the selected gene were frequently present in place of, or in association with, intrachromosomal copies of the gene (Figure 3). The extent of hsr breakdown was variable from cell to cell, possibly depending on the number and the localization of IS copies within individual hsr (Figure 3A,B,C). Hence, under such experimental conditions, dmin and hsr could be present in a single cell, at least transiently. Moreover, the size and the number of dhfr genes present on each extrachromosomal element were heterogeneous from cell to cell, as well as within individual cells in either clone (Figure 3C). The rest of the cells (about 75%) were apparently unaffected (not shown). The similarity of the frequencies of rearranged cells and of cells expressing the gfp protein strongly suggests that breaks occurred in all the cells that received the plasmid and that hsr breakage efficiently lead to dmin formation. In good agreement with this hypothesis, control experiments indicated that 3% at most of the cells transfected with pcDL-SR296 (or untreated) exhibited a rearranged hsr (Table 1) and/or dmin.

When transfected cell populations were grown for a few more days in medium containing MTX, some cells acquired a large number of dmin (Figure 3D). In striking contrast, cells containing extrachromosomal elements were no longer observed after around 2 weeks of growth in the absence of MTX (not shown). Taken together, these observations strongly suggest that the extrachromosomal elements formed upon hsr breakdown segregate unequally at mitosis, as do classical dmin.

I-Sce1 induced DSBs generate several types of chromosomal rearrangements, including inversions

The results presented above indicate that conversion of hsr into dmin represents the majority of the rearrangements occurring in these experimental conditions (Table 1). However, they also disclose the formation of other types of chromosomal rearrangements. Among them, we observed inversions involving the amplified segment from chromosome 2p (Figure 4A). Indeed, such inversions are simply accounted for by the induction of breaks at the parental IS (the one present in cells of line 112) and at an extra-copy of IS, followed by reintegration within the broken chromosome arm of the excised megabase-long linear molecule in an inverted orientation. Inversions looking like the one shown in Figure 4A represented about 9% of the rearrangements involving the hsr. This frequency is certainly underestimated because inversions occurring between two extra-copies of IS escape detection by the cytogenetic analysis performed here.

In addition, ring chromosomes and translocations were also observed at low frequencies (Figure 4B). One of the breakpoints that triggers these rearrangements occurred on the long arm of chromosome 2 or another chromosome, and thus does not rely on an IS. Indeed, the constant presence of MTX in the culture medium could be responsible for this low background of Isce1-independent breaks.

Fragile site activation in clones 112-MTX1 and 112-MTX2 also triggers formation of dmin and chromosomal rearrangements

We determined whether treatment of hsr-containing cells by fragile site activators could induce the same pattern of rearrangements as I-Sce-I expression in IS containing cells. Clones 112-MTX1 and 112-MTX2 were treated by hypoxia, or by aphidicolin or actinomycin D, two drugs that efficiently activate fragile sites in these cells, in conditions giving rise to 5-10 breaks per metaphase plate. The treatments were repeated up to four times on the same cell population. Twenty-four to 72 h after the last treatment, we determined whether the initial hsr was rearranged. The two clones gave comparable results and those obtained for clone 112-MTX1 are presented in Table 2. Compared to control experiments, obvious rearrangements of the hsr were observed in around 25% of the cells upon repeated treatment by hypoxia or either drug. Like the experiments with I-Sce1 described above, dmin and different types of chromosomal rearrangements were generated (Figure 5). In 17 to 20% of the cells treated with either drug and in 10% of the cells submitted to oxygen deprivation, extra-copies were found on extra-chromosomal elements (Figure 5A). Again, a striking heterogeneity was observed both in the size of the dmin and in the number of dhfr copies remaining within the hsr after treatment. When the treated populations were grown for a few more generations in the presence of MTX, cells containing large numbers of dmin could also be observed (Figure 5B). In striking contrast, when MTX was removed from the culture media, cells with dmin rapidly disappeared. Thus, these dmin behave like dmin formed upon I-Sce1-induced DSBs.

In 10, 11 or 18% of the cells challenged by actinomycin D, aphidicolin or hypoxia, respectively, additional chromosomal rearrangements involving the hsr were observed (Table 2). In about 7 to 10% of treated cells, all or a fraction of the amplified copies were translocated to other chromosomes (Figure 5C). We also observed ring chromosomes at low frequency (3%) (Figure 5D) and dicentric chromosomes (Figure 5E). While the latter type of rearrangement was relatively infrequent with drug treatments, dicentric chromosomes were observed in 12% of hypoxia-treated cells (Table 2). In contrast to the results obtained in cell populations transfected with plasmid pI-SceI, inversions taking place within chromosome 2p were not observed among more than 600 metaphase plates analysed after fragile site induction.

Fragile site activation in clones recovered from parental GMA32 cells, amplified for the dhfr or the mdr1 genes, generates extra-chromosomal elements behaving like classical dmin

In order to verify that the genome instability detected in cells challenged with DNA damaging drugs is not influenced by the presence of IS within the hsr, we repeated the experiments described above with dhfr amplified cells derived from the parental GMA32 cell line. Fragile sites were activated in cells of two clones, GMA32-MTX1 and GMA32-MTX2 (Coquelle et al., 1997), exhibiting an hsr with the cytogenetic features of BFB cycles. Results were very similar to those described above for clones 112-MTX1 and 112-MTX2. In particular, cells with dmin were observed in 18 to 25% of the cells treated with actinomycin D or aphidicolin, and in some 12% of the cells challenged by hypoxia. This definitively establishes that the presence of the plasmid did not influence the outcomes of fragile site activation in our previous experiments.

We then extended our analysis to another locus by studying two clones amplified for the mdr1 gene, which also exhibited the cytogenetic features of the BFB cycles (Coquelle et al., 1997). We chose to study clones recovered upon adriamycin (GMA32-ADR1) or actinomycin D (GMA32-AMD1) selection. The cells were challenged with aphidicolin or MTX, two drugs that are not detoxified through the mdr1 pathway and still induce fragile sites in these lines. In both clones, hsr destabilization generated dmin, at a frequency close to that observed in the case of cells amplified for the dhfr gene analysed above (not shown). Thus, inter-conversion of hsr into dmin upon fragile site activation is not restricted to the dhfr locus.

Discussion

Site-specific DSBs and chromosomal rearrangements

Accurate repair of DNA damage, such as DSB, is of crucial importance for the maintenance of genome stability. Recently, model systems based on I-SceI targeted DSBs have been developed, which make it possible to evaluate the contribution of DSBs to the generation of chromosomal rearrangements observed in tumor cells. Imperfect repair of a DSB induced within a cell containing a single copy of IS generates interstitial deletions or triggers BFB cycles (Pipiras et al., 1998). More recent work demonstrated that I-SceI-induced DSBs, initiated from two IS lying on different chromosomes, promotes reciprocal translocations (Richardson and Jasin, 2000). Here we show that multiple I-SceI-induced DSBs on chromosome 2p generated paracentric inversions, easily identified by the relative positions of characteristic DAPI bands that mark the amplified chromosome arm and of the extra-copies of the dhfr gene. In contrast, pericentric inversions, centric ring chromosomes and translocations, the formation of which requires DSBs on both arms of a chromosome or on two different chromosomes, occur infrequently under these conditions. We propose that these latter rearrangements result from a background level of breakage taking place outside the hsr, which probably is a consequence of the presence of MTX in the culture medium. As expected, these rearrangements were more frequently observed in cells treated under conditions that efficiently activate fragile sites, while paracentric inversions affecting chromosome 2p were not observed.

Mechanisms of dmin formation and inter-conversion of amplified structures

We show that in cells bearing amplified copies of the dhfr gene within an hsr, treatments generating DSB induce various rearrangements, including hsr disintegration and formation of dmin. Drugs and hypoxia were less prone than I-Sce1 to promote dmin formation. Indeed, the localization of the breaks induced by each treatment-specifically within the hsr or both within the hsr and the rest of the genome-determine the relative frequency and the pattern of genome reshuffling in each case. We also show that growth conditions strongly select for some types of cells and/or amplified structures during clonal expansion. Cells containing dmin are no longer observed upon removal of MTX from the culture medium for a few generations. This is in good agreement with the general instability of acentric extra-chromosomal elements. In striking contrast, when treated populations are maintained in selective medium, numerous cells contain a number of dmin which is far higher than what could be expected from the number of breaks induced by I-Scel expression or fragile site activation. The observation of such cells strongly suggests that the extrachromosomal elements formed upon hsr degradation segregate unequally at mitosis as do classical dmin. As observed previously (Coquelle et al., 1998), we also observed here that when a cell contains numerous dmin, they are generally small, homogeneous in size and bear only a few copies of the selected gene. This suggests that large extra-chromosomal elements are rapidly lost, possibly because they are particularly unstable. We propose that, in the presence of MTX, some cells that accumulate extra-copies of the dhfr gene through uneven segregation of a sub-population of small and more stable elements rapidly overgrow the population.

Finally, we confirmed that dmin can be formed through hsr breakdown by studying the effects of fragile site activation on hsr containing amplified mdr1 genes, a conclusion also supported by results recently obtained in human cells (Singer et al., 2000). Thus, the phenomenon described here is limited neither to a particular locus nor to Chinese hamster cells.

Previous studies have shown that, at early stages of the amplification process, hsr and dmin bearing extra-copies of the ampd2 or mdr1 genes result from independent amplification mechanisms. Indeed, while the operation of the BFB cycles is accompanied by heavy rearrangements of the amplified chromosome arm (Coquelle et al., 1997), such rearrangements are not present in early mutants that exhibit extra-copies of the ampd2 or the mdr1 genes on dmin. Moreover, DNA damaging agents that activate common fragile sites trigger BFB cycles at both loci while they fail to induce the formation of dmin. This unambiguously establishes the existence of an initial mechanism of dmin formation, that does not rely on DSB at fragile sites (Coquelle et al., 1997; Toledo et al., 1992, 1993). We show here that dmin can also be formed through hsr breakdown, which reveals a second pathway leading to dmin formation (Figure 1). A third mechanism, relying on chromosome breakage occurring within replication eyes, was proposed a number of years ago (Ruiz and Wahl, 1988; Windle et al., 1991). However, all observations reported by this group, including the palindromic organization of the repeated sequences on extrachromosomal elements, are equally consistent with the operation of a mechanism generating dmin from hsr breakdown, the existence of which is unambiguously established by the results presented here. Indeed, the BFB cycle mechanism generates intrachromosomal copies arranged as large palindromes and subsequent breakdown of such hsr is expected to give rise to dmin containing inverted repeats.

Fragile sites have also been involved in a growing number of chromosomal rearrangements found in human cancer, such as deletions of tumor suppressor genes, translocations and oncogene amplifications (Fang et al., 2001; Krummel et al., 2000; Le Beau, 1986; Perucca-Lostanlen et al., 1997; Richards, 2001; Smith et al., 1998; Sutherland et al., 1998; Yunis, 1983). In vivo, stresses resulting from variations in the tumor microenvironment may be important for fragile site activation. We show here that hypoxia induces hsr breakdown, formation of dmin and other chromosomal rearrangements. Thus, stresses activating fragile sites might directly drive most genome rearrangements involved in tumor progression.

Materials and methods

Selection of amplified mutants

The wild type GMA32 cell line, line 112 and respective culture media have been previously described (Pipiras et al., 1998). The method used to observe cells at early stages of gene amplification is derived from Smith et al. (1990). Sub-populations of 50 cells were plated in separate dishes and grown in regular medium for about 12 generations. Cells were then plated in selective medium containing actinomycin D (100 ng/ml) or MTX (9 or 18 ng/ml) for mdr1 and dhfr selection, respectively. Drug-resistant colonies were recovered and expanded for a few cell generations before cytogenetic analysis. Exponentially growing cells were treated with nocodazole (10 M) for 2-4 h, then spread on slides as previously described (Toledo et al., 1992) and used for FISH analysis.

Induction of I-SceI breaks

The experiments were performed with two independent clones, 112-MTX1 and 112-MTX2, exhibiting a co-amplification of IS and the dhfr gene within an hsr. Cells of both populations were transfected with a plasmid coding for the I-SceI nuclease (pI-SceI), or with an empty vector (pcDL-SR296) as previously described (Pipiras et al., 1998). Transfection efficiencies were monitored by co-transfecting plasmid pEGFP-N1 (Clonetech), a plasmid coding for the green fluorescent protein (gfp). Fluorescent positive cells were scored 48 h after transfection. Transfection experiments were performed using lipofectamine^Ô reagent (Life Technologies). Cells were treated for 5 h as recommended by the supplier. Five days after transfection, the cells were recovered and studied by FISH.

Induction of fragile sites and FISH analysis

Exponentially growing cells (5.10⁵ cells per 10 cm diameter petri dishes) amplified for the dhfr gene were continuously grown in the presence of 18 ng/ml MTX. They were submitted to anoxia for 5 h as previously described (Coquelle et al., 1998) or challenged for 18 h in medium supplemented with actinomycin D (100 ng/ml) or aphidicolin (0.4 or 0.6 g/ml) and allowed to recover for 24-48 h in MTX-containing medium. Treatments were repeated four times as previously described (Coquelle et al., 1998). Six hours after the last treatment, the cells were treated with nocodazole (10 M) for 2-4 h, then spread on slides and used for FISH studies. Cosmids KP 454 and KZ 381 (kind gifts of J Hamlin) were used to probe the dhfr locus and plasmid pI-SceI to probe the plasmid. The hybridization conditions and the revelation protocols were previously described (Coquelle et al., 1998; Pipiras et al., 1998).

Acknowledgements

We thank Drs M Weiss and G Buttin for careful reading of the manuscript and helpful discussion, B Dujon for providing the I-SceI gene and the protein recognition sequence, J Hamlin for the cosmids of the dhfr locus. This work was supported in part by the Association pour la Recherche sur le Cancer, the Fondation de France, the Ligue Nationale Française contre le Cancer (Comité de Paris). A Coquelle is supported by a fellowship of the Ligue Nationale Contre le Cancer and L Rozier is a Fellow of the MENRT.

Balaban-Mallenbaum G, Grove G, Gilbert FW. (1981). Cancer Genet. Cytogenet., 2: 339-348.

Brison O. (1993). Biochem. Biophys. Acta, 1155: 25-41.

Brizel DM, Scully SP, Harrelson JM, Layfield LJ, Bean JM, Prosnitz LR, Dewhirst MW. (1996). Cancer Res., 56: 941-943. MEDLINE

Carroll SM, DeRose ML, Gaudray P, Moore CM, Needham-Vandevanter DR, Von Hoff DD, Wahl GM. (1988). Mol. Cell. Biol., 8: 1525-1533. MEDLINE

Coquelle A, Pipiras E, Toledo F, Buttin G, Debatisse M. (1997). Cell, 89: 215-225. MEDLINE

Coquelle A, Toledo F, Stern S, Bieth A, Debatisse M. (1998). Mol. Cell, 2: 259-265. MEDLINE

Dujon B. (1989). Gene, 82: 91-114.

Fang JM, Arlt MF, Burgess AC, Dagenais SL, Beer DG, Glover TW. (2001). Genes Chromosomes Cancer, 30: 292-298.

Gisselsson D, Pettersson L, Hoglund M, Heidenblad M, Gorunova L, Wiegant J, Mertens F, Dal Cin P, Mitelman F, Mandahl N. (2000). Proc. Natl. Acad. Sci. USA, 97: 5357-5362. MEDLINE

Graeber TG, Osmanian C, Jacks T, Housman DE, Koch CJ, Lowe SW, Giaccia AJ. (1996). Nature, 379: 88-91. MEDLINE

Hellman A, Zlotorynski E, Scherer SW, Cheung J, Vincent JB, Smith DI, Trakhtenbrot L, Kerem B. (2002). Cancer Cell, 1: 89-97.

Jasin M. (2000). Cancer Invest., 18: 78-86. MEDLINE

Kinzler KW, Vogelstein B. (1996). Nature, 379: 19-20. MEDLINE

Knudson AG. (2000). Annu. Rev. Genet., 34: 1-19. Article MEDLINE

Krummel KA, Roberts LR, Kawakami M, Glover TW, Smith DI. (2000). Genomics, 69: 37-46. Article MEDLINE

Le Beau MM. (1986). Blood, 67: 849-858.

Luk CK, Veinot-Drebot L, Tjan E, Tannock IF. (1990). J. Natl. Cancer Inst., 82: 684-692.

Ma C, Martin S, Trask B, Hamlin JL. (1993). Genes Dev., 7: 605-620. MEDLINE

McClintock B. (1942). Proc. Natl. Acad. Sci. USA, 28: 458-463.

Nowell PC, Rowley JD, Knudson AG. (1998). Nat. Med., 4: 1107-1111.

Perucca-Lostanlen D, Hecht BK, Courseaux A, Grosgeorge J, Hecht F, Gaudray P. (1997). Cytogenet. Cell Genet., 79: 88-91. MEDLINE

Pipiras E, Coquelle A, Bieth A, Debatisse M. (1998). EMBO J., 17: 325-333. Article MEDLINE

Rice GC, Hoy C, Schimke RT. (1986). Proc. Natl. Acad. Sci. USA, 83: 5978-5982.

Richards RI. (2001). Trends Genet., 17: 339-345. Article MEDLINE

Richardson C, Jasin M. (2000). Nature, 405: 697-700. Article MEDLINE

Rouet P, Smih F, Jasin M. (1994). Mol. Cell. Biol., 14: 8096-8106. MEDLINE

Ruiz JC, Wahl GM. (1988). Mol. Cell. Biol., 8: 4302-4313. MEDLINE

Saunders WS, Shuster M, Huang X, Gharaibeh B, Enyenihi AH, Petersen I, Gollin SM. (2000). Proc. Natl. Acad. Sci. USA, 97: 303-308. Article MEDLINE

Schwab M. (1999). Semin. Cancer. Biol., 9: 319-325. Article MEDLINE

Singer MJ, Mesner LD, Friedman CL, Trask BJ, Hamlin JL. (2000). Proc. Natl. Acad. Sci. USA, 97: 7921-7926. MEDLINE

Smith DI, Huang H, Wang L. (1998). Int. J. Oncol., 12: 187-196.

Smith KA, Gorman PA, Stark MB, Groves RP, Stark GR. (1990). Cell, 63: 1219-1227. MEDLINE

Smith KA, Stark MB, Gorman PA, Stark GR. (1992). Proc. Natl. Acad. Sci. USA, 89: 5427-5431. MEDLINE

Sutherland GR, Baker E, Richards RI. (1998). Trends Genet., 14: 501-506. Article MEDLINE

Toledo F, Buttin G, Debatisse M. (1993). Curr. Biol., 3: 255-264.

Toledo F, LeRoscouet D, Buttin G, Debatisse M. (1992). EMBO J., 11: 2665-2673. MEDLINE

Trask BJ, Hamlin JL. (1989). Genes Dev., 3: 1913-1925. MEDLINE

Windle B, Draper BW, Yin Y, O'Gorman S, Wahl GM. (1991). Genes Dev., 5: 160-174. MEDLINE

Yunis JJ. (1983). Science, 221: 227-236.

Figure 1 Initial and secondary mechanisms of formation of dmin. Two homologue chromosomes are represented (light and dark blue). Symbols on the chromosomes are; arrowheads: telomeres, blue circles: centromeres, green sequence: sequence to be amplified, red sequence: distal marker. M: mitosis, R: replication, LO: looping-out, DSB: double-strand break, Red cell: parental cell in G1 phase. Pink and yellow pathways: initial mechanism relying on looping out of a sequence containing the selected gene. Yellow pathway: looping out of the sequence to be amplified in a cell in G1 phase. After one step of replication a dmin is formed. Uneven segregation of the extrachromosomal acentric molecules at mitosis during subsequent cell cycles leads to higher levels of amplification. The cells of the resulting clone are characterized by an interstitial deletion, corresponding to the excised sequence, on one homologue. Pink pathway: looping out in a cell in phase G2. One of the four chromatids of the cell is deleted. After mitosis, one daughter receives two normal chromosomes while the other one receives a normal and a deleted chromosome. In half the cases, the acentric molecule segregates within a cell with two normal chromosomes (left) which later gives rise to an amplified clone devoid of chromosomal abnormalities. In the other cases, the acentric molecule segregates within a cell with one normal and one deleted chromosome (right), a situation indistinguishable from the one depicted in the yellow pathway. The daughter cells that do not receive an extrachromosomal molecule are counter-selected in selective medium (black line). Blue pathway: mechanism relying on hsr breakdown. A DSB induces the BFB cycle mechanism, leading to intrachromosomal in loco amplification. Breakage of the hsr and subsequent formation of circular acentric molecules give rise to clones with dmin, the cells of which have a deletion of the distal part of one homologue. Until healing, this chromosome is unstable

Figure 2 Co-amplification of the I-SceI recognition sequence (IS) and the dhfr gene by BFB cycles in MTX-resistant mutants derived from line 112. (A) Model for the formation of an hsr containing both dhfr genes and IS. Red: dhfr gene. Yellow: IS. Dark blue: dark DAPI band specific to Chinese hamster chromosome 2p. Dotted arrow: multi-step formation of dhfr-containing hsr. Green arrows: breaks at fragile sites distal and proximal to the dhfr gene. According to the BFB cycle mechanism, a break at a fragile site distal to the dhfr gene initiates the process when followed by fusion of the broken sister chromatids, bridge formation and further break at mitosis. (2), (3) models for the structure of chromosomes exemplified in B2 and B3 respectively. (B) Examples of hsr in cells of clone 112-MTX1. (1), (2), (3): Left: DAPI staining. Arrowheads point to the duplicated 2p arm dark DAPI band, which indicates that the initiating break occurred distal to this band. Right: FISH with probes for the dhfr gene (1; revealed by rhodamine) or for the plasmid containing IS (2, 3; revealed by fluorescein). In cells of these clones, as shown (1), numerous dhfr extra-copies were present and distributed up to the extremity of the amplified chromosome 2p. (2) and (3) point to the high level of heterogeneity observed from cell to cell within each clone, both for the number and organization of IS along the hsr

Figure 3 Examples of I-SceI-induced hsr degradation generating extra chromosomal elements. (A) Formation of a chromosome fragment. Top: DAPI staining. Note the presence of a faint link between the fragment and the rest of the chromosome. Bottom: FISH with a probe for the dhfr gene (revealed by rhodamine). (B, C, D) Examples of cells containing extra chromosomal elements resulting from the degradation of the initial hsr. FISH with a probe for the dhfr gene (revealed by rhodamine). (B, C) Arrowheads point to amplified chromosome 2p. Note the variable number of extra-copies of the dhfr gene remaining on this chromosome. (C) Arrows point to extra chromosomal elements heterogeneous in size and in dhfr gene copy number within the same cell. (D) Cell exhibiting a very high number of dmin

Figure 4 Examples of I-SceI-induced hsr rearrangements. (A) Inversion within amplified chromosome 2p. Left: DAPI and FISH with a probe for the dhfr gene (revealed by rhodamine). Note the position of the duplicated DAPI band (arrowheads) relative to the cluster of dhfr gene copies. Right: model for the formation of this inversion. Red: dhfr gene, dark blue: dark DAPI band specific to Chinese Hamster chromosome 2p, green arrows: I-SceI induced breaks. In this example, the location of the breaks left no dhfr gene between the double DAPI band and the chromosome end. (B) Single color FISH with a probe for the dfhr gene (red) showing: (1) a translocation of the amplified copies on an unidentified acrocentric chromosome, (2) a ring chromosome bearing dhfr amplified genes

Figure 5 Examples of formation of dmin and hsr rearrangements upon treatment by hypoxia. FISH with a probe for the dhfr gene (revealed by fluorescein). Arrowheads point to duplicated dark DAPI bands specific for dhfr amplified mutants. (A, B, C) Small arrows point to the normal chromosome 2. (A) Example of partial hsr breakdown, large arrows point to three extra chromosomal elements present in this metaphase plate. (B) Example of a cell presenting extensive hsr breakdown associated with the presence of a high number of dmin. (C) Translocation of part of the dhfr extra-copies on a chromosome 1, leaving dhfr copies in the initial hsr. (D) Example of ring chromosome involving the amplified chromosome 2. Top: DAPI staining, bottom: FISH. Note the conservation of the typical organization of amplified dhfr gene copies relative to the duplicated dark DAPI band. (E) Example of a symmetric dicentric chromosome involving amplified chromosomes 2. Top: DAPI staining, bottom: FISH. The typical structure of dhfr amplified mutants is also maintained

Table 1 Effects of I-SceI transfection on the stability of hsr containing amplified dhfr genes

Table 2 Effects of treatments activating fragile sites on the stability of hsr containing amplified dhfr genes