DNA Amplification by Breakage/Fusion/Bridge Cycles Initiated by Spontaneous Telomere Loss in a Human Cancer Cell Line

Abstract

The development of genomic instability is an important step in generating the multiple genetic changes required for cancer. One consequence of genomic instability is the overexpression of oncogenes due to gene amplification. One mechanism for gene amplification is the breakage/fusion/bridge (B/F/B) cycle that involves the repeated fusion and breakage of chromosomes following the loss of a telomere. B/F/B cycles have been associated with low-copy gene amplification in human cancer cells, and have been proposed to be an initiating event in high-copy gene amplification. We have found that spontaneous telomere loss on a marker chromosome 16 in a human tumor cell line results in sister chromatid fusion and prolonged periods of chromosome instability. The high rate of anaphase bridges involving chromosome 16 demonstrates that this instability results from B/F/B cycles. The amplification of subtelomeric DNA on the marker chromosome provides conclusive evidence that B/F/B cycles initiated by spontaneous telomere loss are a mechanism for gene amplification in human cancer cells.

Introduction

Genomic instability is an important step in generating the multiple genetic changes required for cancer [1–3]. Genomic instability can result in both the loss of tumor suppressor genes through deletions or chromosome loss, or the overexpression of oncogenes through gene amplification [4,5]. Gene amplification can occur through a number of different mechanisms. In human tumor cells, high-copy gene amplification most often involves the formation of doubleminute (DM) chromosomes [6,7]. Gene amplification can also result from the breakage/fusion/bridge (B/F/B) cycles, which were first described in maize [8]. B/F/B cycles are initiated when broken chromosomes fuse and then break during anaphase when the two centromeres are pulled in opposite directions. The B/F/B cycle is then continued in the next cell cycle when sister chromatids fuse following DNA replication. B/F/B cycles have been found to be the primary mechanism for gene amplification in hamster cells [9–12]; however, the importance of B/F/B cycles in genomic instability in human cancer is unclear.

The ends of chromosomes, called telomeres, play an important role in maintaining chromosome stability and preventing chromosome fusion. Telomeres are composed of a 6-bp repeat sequence and associated proteins that form a cap that protects chromosome ends [13,14]. Telomeric repeat sequences are added on by the enzyme telomerase, which compensates for the loss of the ends of chromosomes during cell division. In humans, telomerase activity is expressed in germ line cells, but not in most somatic cells [15]. As a result, in somatic cells, telomeres continue to shorten with each cell division, which has been proposed to be the mechanism for mediating cell senescence [16]. Consistent with this hypothesis, expression of the gene for the catalytic subunit of telomerase is capable of extending the lifespan of primary human fibroblasts [17]. Thus, to overcome telomere shortening, immortalized cells invariably regain the ability to maintain telomeres, either through the activation of telomerase activity [16,18] or through the activation of an alternative mechanism that appears to involve recombination [19,20]. The importance of telomeres in preventing chromosome fusion is illustrated by the massive increase in chromosome fusion observed in somatic cells that fail to senesce and therefore continue to undergo telomere shortening [18]. Increased chromosome fusion is also seen in cells with mutations in genes that affect telomere function [21–25]. The ability of cells to properly maintain telomeres is therefore a critical factor in maintaining chromosome stability.

Although telomeres must be maintained for cells to become immortal, cancer cells commonly have problems in maintaining telomeres and preventing chromosome fusion. Many cancer cells demonstrate a high frequency of telomere associations involving chromosomes joined at or near their telomeres, which has been proposed to result from a failure to properly maintain telomere length [15]. Consistent with this observation, many early passage tumor cells in culture demonstrate telomere instability leading to chromosome fusion [26]. A variety of mechanisms may be involved in spontaneous telomere loss. Fluctuations in telomerase activity in some cancer cell lines have been demonstrated to result in global changes in telomere length and increased chromosome fusion [27,28]. However, other cancer cell lines can demonstrate a high frequency of chromosomes without detectable telomeres, which is independent of average telomere length [26,29], suggesting that stochastic mechanisms can also result in telomere loss. Stochastic mechanisms of spontaneous telomere loss could include a failure to replicate telomeres during DNA synthesis or double-strand breaks occurring within or near telomeric repeat sequences, which has been shown to result in telomere loss in mouse ES cells [30].

As first proposed by McClintock [8], the addition of telomeres to the ends of broken chromosomes can promote chromosome stability by preventing or terminating B/F/B cycles. Broken chromosomes can acquire telomeres by a variety of mechanisms. Telomeres can be added on directly to the ends of broken chromosomes by telomerase, as demonstrated in Tetrahymena [31] and yeast [32]. Direct addition of telomeres to the ends of broken chromosomes has been associated with human genetic disease [33–35], and has been observed at the location of double-strand breaks in mouse embryonic stem (ES) cells [36]. However, it is unclear whether telomerase is involved in the direct addition of telomeres in mammalian cells. Telomeres can also be acquired through the translocation of the ends of other chromosomes, as has been observed in human cancer cells [37,38] and in mouse ES cells [30]. Finally, telomeres can be acquired through break-induced replication, as has been observed in yeast [39,40]. In break-induced replication, the broken end of the chromosome invades a region of homology on another chromosome and replicates the end of the chromosome. Whether break-induced replication occurs in mammalian cells is not known.

Evidence that telomere loss and B/F/B cycles play an important role in human cancer is provided by the presence of a high frequency of anaphase bridges in many early passage tumor cells [26] and tumors [41]. In addition, structures consistent with B/F/B cycles have been observed in low-copy gene amplification in some human tumor cells lines [42,43]. B/F/B cycles could also be an early step in high-copy gene amplification. Although human tumor cells in culture selected for gene amplification usually contain DM chromosomes, one clone with low-copy amplification contained structures consistent with B/F/B cycles [44]. When placed under more stringent selection conditions, the amplified genes in this clone converted to DM chromosomes, leading to the proposal that B/F/B cycles are an early step in the formation of DM chromosomes. Consistent with this hypothesis, regions amplified by B/F/B cycles in hamster cells have also been shown to form DM chromosomes [11].

We previously investigated the consequences of spontaneous telomere loss in a human tumor cell line that has a herpes simplex virus thymidine kinase (HSV-tk) selectable marker gene integrated immediately adjacent to a telomere [38]. Loss of the HSV-tk gene was found to be associated with the formation of inverted repeats, large duplications, and prolonged periods of instability in the “marker” chromosome containing the telomeric plasmid sequences. These results suggested that chromosome instability resulting from telomere loss involved sister chromatid fusion and B/F/B cycles. To confirm the presence of B/F/B cycles, we have now analyzed the cells containing the unstable marker chromosome for the presence of anaphase bridges. The extent of amplification of subtelomeric DNA on the marker chromosome was also investigated to determine whether B/F/B cycles initiated by spontaneous telomere loss are a mechanism for gene amplification in human cancer cells.

Materials and Methods

Cell Culture

The EJ-30 cell line (obtained from Dr. William Dewey, University of California, San Francisco) was subcloned from the EJ bladder cell carcinoma cell line, which is also named MGH-U1 [45]. Clones A3 and B3 were isolated from EJ-30 following transfection with the pNCT-tel plasmid, and contain a single copy of the plasmid integrated at the end of a chromosome [38]. After isolation, the clones were expanded to approximately 10⁷ cells prior to Southern blot analysis and isolation of subclones deficient in HSV-tk. HSV-tk-deficient subclones were selected from multiple experiments using medium containing both ganciclovir (50 µM) and G418 (400 µg/ml) to ensure that some portion of the plasmid was present to allow for analysis of the recombination events involved.

Southern Blot Analysis

Genomic DNA purified as previously described [46] was digested with restriction enzymes according to the manufacturer's instructions. For analysis of terminal fragments, the DNA was digested with BAL31 following the manufacturer's recommendations (Promega, Madison, WI), extracted with phenol/chloroform, precipitated, resuspended, and digested with BamHI. Genomic DNA was fractionated by agarose gel electrophoresis using standard protocols, depurinated by treatment with 0.25 M HCl for 30 minutes, and transferred in 0.5 M NaOH onto a charged nylon Hybond-N⁺ membrane (Amersham, Piscataway, NJ) using a vacuum transfer apparatus (Amersham). Prehybridization for 3 hours and hybridization overnight were performed at 658C in 5x SSPE, 5x Denhardt's solution, 0.5% sodium dodecyl sulfate (SDS), and 0.25 mg/ml salmon sperm DNA. Probes were labeled with [α³²P]dCTP (New England Nuclear, Chicago, IL) using a High Prime labeling kit (Roche, Indianapolis, IN). Filters were washed three times in 2x SSPE with 0.1% SDS at room temperature, twice in 1x SSPE with 0.1% SDS at 658C, and twice in 0.1x SSC with 0.1% SDS at 658C.

Chromosome Analysis

For analysis of anaphase cells, cultures were grown as monolayers in chamber slides and fixed with ice-cold methanol for 15 minutes followed by ice-cold methanol: acetone (1:1 vol/vol) for 30 minutes. FISH analysis was performed using Spectrum Green-labeled chromosome 16-specific painting probe (Vysis, Downers Grove, IL) following the manufacturer's protocol. DNA was counterstained with 4,6-diamino-2-phenylindole (DAPI). Anaphase cells were identified by their characteristic chromosomal configurations.

Preparation of metaphase chromosomes and hybridizations were performed as previously described [47,48]. Cosmid clones RT99 (Genbank accession no. AC004653) and 317H7 (Genbank accession no. AC005569) have been mapped to the end of chromosome 16p [49]. These cosmids were isolated from partially digested DNA libraries made from flow-sorted human chromosome 16 [50]. To improve separation of FISH signals, chromosomes were stretched. Cells were treated with colcemid (0.4 µg/ml, 30 minutes), harvested by trypsinisation, swollen with a hypotonic solution of 0.075 M KCl/serum (5:1 vol/vol), and sedimented onto glass slides for 5 minutes at 1000 rpm (Cytospin; Shandon, Pittsburgh, PN). Chromosomes were fixed overnight in methanol/acetic acid (3:1 vol/vol). The BAC probes that were used consisted of GS-121-I4, which is located a maximum of 160 kb from the telomere on 16p, GS-240-G10 for the long arm of chromosome 16 (16q), and BAC GS-121-I4, which is located a maximum of 200 kb from the telomere on 16q [51].

Telomere analysis was performed as previously described [52] using telomere-specific peptide nucleic acid (PNA) probes labeled with Cy3 (Perseptive Biosystems, Foster City, CA).

Results

Chromosome Instability Associated with Spontaneous Telomere Loss in a Human Tumor Cell Line

The relationship between telomere loss and chromosome instability in human cells was investigated using clone B3 of the human EJ-30 bladder cell carcinoma cell line [38]. Clone B3 contains a selectable HSV-tk gene located immediately adjacent to a telomere on chromosome 16 (Figure 1A). Cell clones containing selectable marker genes adjacent to a telomere are established by transfection with linearized plasmids containing telomeric repeat sequences on one end. The integration of the plasmid sequences on the end of a broken chromosome results in a new telomere “seeded” from the telomeric repeat sequences within the plasmid [53–55]. The seeded telomeres on these marker chromosomes are elongated in culture, and their length and dynamics become similar to the other telomeres in the cell [29,56]. In the EJ-30 clones A3 and B3, the terminal BamHI restriction fragments average approximately 7.5 and 9.5 kb, respectively, indicating telomeres that average approximately 3 and 5 kb in length (Figure 1B). Subclones of A3 and B3 selected at random demonstrated variability in the length of the terminal restriction fragments, ranging from 7 to 10 kb in clone B3 and from 9 to 12 kb in clone A3 (Figure 1B). With increasing time in culture, clone B3 showed no change in the length of the seeded telomere, whereas subclones of B3 showed gradual changes in the length of the seeded telomere (Figure 1C). These results are similar to those obtained with other tumor cell lines, in which the length of the seeded telomeres is at equilibrium in the original clones, whereas telomere length in subclones is often not at equilibrium but returns to equilibrium with time in culture [29,56]. Thus, the seeded telomeres in clones of EJ-30 are relatively stable, although some variation in telomere length is observed in different cells in the population.

Figure 1.

Structure of the integrated plasmid sequences and length of seeded telomeres in EJ-30 clones A3 and B3. (A) Clones A3 and B3 contain a single copy of the pNCT-tel plasmid integrated on the end of a chromosome. The locations of the cellular DNA, plasmid vector (amp/ori), neo gene, HSV-tk gene, and telomeric repeat sequences are shown. The location of BamHI restriction sites (Bm) used for Southern blot analysis and the size of the BamHI restriction fragments are shown. (B) Southern blot analysis of BamHI-digested genomic DNA from clones A3 and B3, and subclones 1 through 5 selected at random from each clone. Hybridization was performed with the pNCT-Δ probe, which is identical to pNCT-tel except that it does not contain telomeric repeat sequences. (C) Southern blot analysis of BamHI-digested genomic DNA from clone B3 and its subclones B3-1 and B3-2 after different numbers of cell doublings using the pNCT-Δ plasmid as a probe. The internal BamHI fragments in clones A3 and B3 are both approximately 4.5 kb in length. The terminal fragments containing 4.4 kb of plasmid sequences are heterogeneous in length due to variability in the length of the telomeres in different cells in the population. Molecular size markers consisting of Lambda bacteriophage HindIII restriction fragments are shown.

Selection with ganciclovir for cells that had lost the HSV-tk gene provides a method for following the consequences of spontaneous telomere loss on a specific chromosome without selection for gene amplification. Clones of EJ-30 containing telomeric integration sites showed a high rate of spontaneous loss of the HSV-tk gene (10^-4 events/cell per generation) as compared to clones with interstitial integration sites [38]. In contrast, the seeded telomeres in mouse ES cells are relatively stable and spontaneous loss of the HSV-tk gene is below the level of detection [30]. Thus, the seeded telomeres in tumor cell lines are relatively unstable, consistent with studies indicating telomere instability in tumor cells and tumor cell lines [26,29]. The analysis of four HSV-tk-deficient (HSV-tk^-) subclones of clone B3, G45, G55, G60, and G71 demonstrated the presence of inverted repeats within the plasmid sequences, whereas cytogenetic analysis showed the presence of large duplications on the end of the marker chromosome [30,38]. The heterogeneity in the structure of the marker chromosome in different cells in the population and the continued instability of the marker chromosome in many second-generation subclones demonstrated that this instability continues for many cell generations. The combined results from these experiments strongly indicated that sister chromatid fusion was the first event leading to chromosome instability resulting from spontaneous telomere loss.

Anaphase Bridges Associated with Chromosome Instability Resulting from Spontaneous Telomere Loss

The presence of inverted repeats, large duplications, and prolonged periods of chromosome instability following the loss of a telomere are consistent with a mechanism involving B/F/B cycles (Figure 2). To determine whether B/F/B cycles are involved in the instability in the marker chromosome, we examined subclone G71 for the presence of anaphase bridges, which are considered a hallmark of chromosome fusion resulting from telomere loss [8,26,41]. Anaphase bridges were observed in 10% (21 of 200) of the anaphase cells in subclone G71. In contrast, the parental B3 cell line had significantly fewer bridges in anaphase cells (1 of 200). In addition, hybridization with a chromosome 16-specific probe showed that the anaphase bridges in subclone G71 consisted of chromosome 16 in 9 of 11 anaphase bridges that were analyzed (Figure 3). The presence of a high frequency of anaphase bridges involving chromosome 16 provides compelling evidence that the marker chromosome in subclone G71 is involved in B/F/B cycles.

Figure 2.

The mechanism of gene amplification involving B/F/B cycles. B/F/B cycles are initiated when sister chromatids fuse following the loss of a telomere. The resulting dicentric chromosome forms a bridge during anaphase and breaks again, continuing the cycle until the chromosome obtains a new telomere. Breakage at locations other than the site of fusion results in amplification of sequences in one daughter cell and deletions in the other daughter cell. The distance between the amplified arrays is dependent upon the distance of the break from the site of fusion. The telomeres (gray squares), centromeres (circles), and orientation of the subtelomeric sequences (arrow) are shown.

Figure 3.

Anaphase bridges specific to chromosome 16 in subclone G71. The staining of total chromatin with DAPI (left panel, pseudo-colored orange) is used to identify anaphase cells without (A) and with (B,C) chromosome bridges (arrows). Hybridization with a chromosome 16-specific painting probe (center panel, green) demonstrates that most anaphase bridges in subclone G71 involve chromosome 16. The merged image showing both DAPI staining and hybridization with the chromosome 16-specific probe is also shown (right panel).

DNA Amplification Resulting from B/F/B Cycles Initiated by Spontaneous Telomere Loss

B/F/B cycles can result in amplification of DNA (Figure 2), and amplified regions consistent with B/F/B cycles have been observed in a number of mammalian cells [9–12]. To determine whether B/F/B cycles in clone B3 result in gene amplification, metaphase chromosomes from subclone G71 were hybridized with cosmid RT99 that contains sequences subtelomeric to the plasmid on the end of the marker chromosome (Figure 4Ai). The same metaphase chromosome preparations were also hybridized with a telomeric probe to monitor the presence of a telomere (Figure 4Aii) and chromosome 16-specific probes to identify the marker chromosome (Figure 4, Aiii and iv). A chromosome 16-specific painting probe was used to identify both chromosome 16 homologs (Figure 4Aiii), whereas hybridization with two bacterial artificial chromosome (BAC) clones specific for the p and q arms of chromosome 16 was used to distinguish between the two homologs (Figure 4Aiv). Due to the deletion of approximately 4Mb of DNA from the end of the p arm upon integration of the telomeric plasmid in clone B3, the marker chromosome does not hybridize with the p arm-specific BAC clone GS-121-I4. In many metaphase spreads, the hybridization signal with the RT99 cosmid probe was located internally within the marker chromosome (Figure 4Ai), consistent with previous results demonstrating duplications on the end of chromosome 16 in subclones containing inverted repeats [38]. The lack of a detectable telomere and a rounded end on the p arm in approximately 20% of the cells in subclone G71 were consistent with continued sister chromatid fusion in many cells in the population (Figure 4Aii). In addition, of 72 metaphase spreads analyzed, all but one showed an increase in hybridization intensity of the subtelomeric RT99 probe in the marker chromosome over that observed in the other chromosome 16 homologue. In approximately half of the metaphases examined (35 of 72), the increase in hybridization intensity with RT99 was small and localized in single spots (Figure 4Ai). In most of the other cells in the population (32 of 72), the increase in hybridization was more pronounced and multiple spots were observed (Figure 4B). Only a small fraction of the cells (4 of 72) contained multiple bands typical of the hybridization patterns previously reported for genes amplified by B/F/B cycles in hamster [9–12] or human [42,43] cells. In contrast to the marker chromosome, the corresponding region in the homologous chromosome in all of the metaphase spreads examined appeared normal with no indication of amplification.

Figure 4.

Gene amplification resulting from the spontaneous telomere loss. (A) Metaphase chromosomes from subclone G71 hybridized with (i) a subtelomeric cosmid probe RT99 and counterstained with propidium iodide; (ii) a telomere-specific PNA probe and counterstained with DAPI; (iii) a chromosome 16-specific painting probe and counterstained with propidium iodide; and (iv) both a chromosome 16q-specific BAC clone GS-240-G10 (red) and a chromosome 16p-specific BAC clone GS-121-I4 (green). The marker chromosome is identified by the absence of hybridization with the GS-121-I4 BAC clone due to the terminal deletion associated with integration of the pNCT-tel plasmid. (B) Two additional metaphase chromosomes of subclone G71 demonstrating increased hybridization with the subtelomeric cosmid probe RT99 and variability in the size of the fragment joined onto the end of the marker chromosome. (C) Three metaphase spreads containing stretched chromosomes were hybridized with fluorescein-labeled cosmid RT99 (green) located immediately adjacent to the telomere and rhodamine-labeled cosmid 317H7 (red) located 1 Mb from the telomere. Overlapping RT99 and 317H7 hybridization signals appear yellow. Chromosomes were counterstained with DAPI (blue). The inserts are magnified views of the amplified regions.

To investigate the structure of the amplified region at higher resolution, stretched metaphase chromosomes from subclone G71 were hybridized simultaneously with a fluorescein-labeled cosmid RT99 located adjacent to the integration site, and a rhodamine-labeled cosmid 317H7 located 1 Mb from the integration site. Multiple hybridization signals involving both cosmid probes were commonly observed, demonstrating amplification of the subtelomeric region (Figure 4C). However, the amplified arrays did not consist of alternating signals from the two cosmids, which would result if the breaks occurred outside of the amplified region (see Figure 2). Instead, most hybridization signals involved cosmid RT99 adjacent to the site of fusion, with only occasional signals involving the more distal cosmid 317H7. This observation demonstrates that most breaks occurred within 1 Mb of the site of fusion and, as a result, the sequences in clone 317H7 are often excluded from the amplified region. The presence of as many as six separate hybridization signals with cosmid RT99 in some metaphase spreads demonstrated that the marker chromosome had been involved in at least three B/F/B cycles.

Discussion

The results presented with EJ-30 [38] and from other studies involving human tumor cells in culture [26–29] demonstrate that human tumor cells often have a high rate of spontaneous telomere loss. A variety of factors could contribute to this apparent inability to properly maintain functional telomeres in the EJ-30 tumor cell line. One possible mechanism is a periodic fluctuation in the level of telomerase, which has been reported to result in a dramatic shortening of telomeres in some cells in the population in some tumor cell lines [27,28]. Some fluctuation in telomere length is seen on the marker chromosome in clones of EJ-30 (Figure 1). However, analysis of telomeres in the HSV-tk-deficient subclones using PNA probes showed that the telomeres in these cells are not significantly shorter in length than the parental cell lines (data not shown). Therefore, the loss of the telomere on the marker chromosome in B3 results from either a transient shortening of all telomeres, or stochastic events occur that result in the complete loss of individual telomeres. Regardless of the mechanism, it is clear from our results that telomere loss can have a dramatic influence on chromosome stability.

Previous studies on the mechanism of gene amplification in mammalian cells have primarily focused on the analysis of the structure of highly amplified regions. In the present study, we have analyzed the mechanism of gene amplification by following the fate of a single chromosome after the loss of a telomere. As a result, we have been able to observe the early steps involved in gene amplification, providing several valuable insights into the role of B/F/B cycles in this process. One important insight is that when a telomere is not added to the end, the first event resulting from telomere loss is often sister chromatid fusion. All four of the HSV-tk^- subclones of B3 that the telomere, had contained inverted repeats involving the plasmid sequences [30,38]. Similar inverted repeats were also seen in four HSV-tk^- mouse ES cell subclones in which telomere loss was induced with the I-SceI endonuclease, although in the ES cells the addition of a telomere at the site of the break was the most common event observed [30]. The instability in the marker chromosome in all of the mouse and human subclones containing inverted repeats [30,38] also demonstrates that sister chromatid fusion is often followed by B/F/B cycles. Sister chromatid fusion is also an important mechanism in perpetuating B/F/B cycles, as demonstrated by the cytogenetic analysis showing the apparent fusion of the ends of sister chromatids lacking a telomere in subsequent cell cycles (Figure 4A).

The demonstration in our study of gene amplification and anaphase bridges in the same chromosome proves conclusively that B/F/B cycles can lead to gene amplification in human cancer cells. Previous studies have shown an increased rate of anaphase bridges in human cancer cells [26,57] and the presence of amplified regions consistent with B/F/B cycles [42–44]. However, the experimental systems used in these studies do not allow for the analysis of both anaphase bridges and gene amplification in the same cell line. The presence of bridges involving chromosome 16 in 10% of the anaphase cells in subclone G71 also demonstrates that B/F/B cycles can last for many generations after the initial sister chromatid fusion. This observation is consistent with the lack of detectable telomeres on the end of the marker chromosome in approximately 20% of the cells in subclone G71 [38]. In contrast, a similar study in mouse ES cells found that B/F/B cycles were generally much shorter in duration, with the marker chromosome acquiring a telomere through the translocation of fragments from other chromosomes [30]. In some instances, telomere acquisition in subclone G71 was also accompanied by translocation of fragments from other chromosomes [38]. However, in some cells, new telomeres were observed on the marker chromosome without detectable translocations, indicating that either microtranslocations or direct addition of telomeric repeats was involved. Both microtranslocations [37] and direct addition of telomeres [33–36] have been observed in human cells.

A third important observation in the present study is the small size of the amplified regions in subclone G71. Previous studies have observed that anaphase bridges usually break within preexisting fragile sites [12,43]. However, the results presented here demonstrate a tendency to break within 1 Mb of the original site of fusion. This breakage near the site of fusion could be due to the creation of a new fragile site caused by the presence of the large inverted repeats, which have been shown to be highly unstable [58–60]. A tendency to break near the site of fusion would also explain why the marker chromosome appears relatively “normal” in many cells in the population in subclone G71 despite having undergone one or more sister chromatid fusions [38]. As a result, the typical ladders seen in genes amplified by B/F/B cycles in hamster cells are uncommon in subclone G71. However, this could also be due to the lack of selection for gene amplification in our system, because the genes that are amplified are often far from the initial site of sister chromatid fusion. Regardless, it is clear from our studies that B/F/B cycles can result in the amplification of relatively small regions that are similar in size to the amplified regions commonly seen in human cancer cells [4].

Another interesting observation made in our study is the relatively modest amount of amplification that was observed despite the prolonged B/F/B cycles involving the marker chromosome. Despite the fact that there was no selection for gene amplification, the continuation of B/F/B cycles for more than 20 generations in many cells in the population would provide ample opportunity for extensive gene amplification. These results suggest that B/F/B cycles are not an efficient mechanism for generating high-copy gene amplification in human cells, which could explain why highly amplified genes in human cells are not found in structures typically associated with B/F/B cycles [6,7]. Despite this apparent inefficiency in generating high-copy DNA amplification, B/F/B cycles may be important in gene amplification in human cancer. Many cancer cell lines and early passage cancer cells in culture have high rates of telomere associations [15] and anaphase bridges associated with telomere instability [26,57], suggesting a high rate of chromosome fusion. In fact, low-copy amplified arrays typical of B/F/B cycles have been observed in some human cancer cell lines [42,43]. The presence of inverted repeats in amplified regions in human cancer cells [61] would also suggest that B/F/B cycles can play a role as an early event in the generation of highly amplified genes. Although most highly amplified genes in human cancer cells are located on DM chromosomes, this amplification may have been initiated by B/F/B cycles in some instances. Genes amplified initially through B/F/B cycles have been demonstrated to convert to DM chromosomes [11,44], which would be consistent with our observation that the region containing the inverted repeats appears to form a new fragile site. Cells that convert to DM chromosomes appear to have a selective advantage under more stringent selection conditions [44], possibly because DM chromosomes are a more efficient mechanism for generating high-copy gene amplification than B/F/B cycles.

B/F/B cycles can also lead to chromosome instability by promoting recombination with other chromosomes, which can promote nonreciprocal translocations, transfer of the amplified genes to other chromosomes, and complex chromosome rearrangements [11,38]. The types of translocations and chromosome rearrangements seen in the human tumor cells [38] and mouse ES cells [30] in our studies are typical of the rearrangements seen in tumors in telomerase-deficient mice [62] and human tumors [63]. B/F/B cycles resulting from spontaneous telomere loss are therefore likely be an important mechanism for generating both gene amplification and the chromosome rearrangements commonly found in human cancer.