Abstract
Telomere loss was produced during development of Drosophila melanogaster by breakage of an induced dicentric chromosome. The most prominent outcome of this event is cell death through Chk2 and Chk1 controlled p53-dependent apoptotic pathways. A third p53-independent apoptotic pathway is additionally utilized when telomere loss is accompanied by the generation of significant aneuploidy. In spite of these three lines of defense against the proliferation of cells with damaged genomes a small fraction of cells that have lost a telomere escape apoptosis and divide repeatedly. Evasion of apoptosis is accompanied by the accumulation of karyotypic abnormalites that often typify cancer cells, including end-to-end chromosome fusions, anaphase bridges, aneuploidy, and polyploidy. There was clear evidence of bridge–breakage–fusion cycles, and surprisingly, chromosome segments without centromeres could persist and accumulate to high-copy number. Cells manifesting these signs of genomic instability were much more frequent when the apoptotic mechanisms were crippled. We conclude that loss of a single telomere is sufficient to generate at least two phenotypes of early cancer cells: genomic instability that involves multiple chromosomes and aneuploidy. This aneuploidy may facilitate the continued escape of such cells from the normal checkpoint mechanisms.
TELOMERES are nucleic acid–protein complexes that protect the ends of linear chromosomes from end-to-end chromosome fusions, referred to as the capping function, and that solve the end-replication problem (Blackburn 2001). Most eukaryotes have simple highly repetitive sequences at chromosome termini (Stewart and Weinberg 2006) and have evolved sequence-specific-binding proteins that recognize these terminal repeats and facilitate proper telomere function. One of these proteins, telomerase, is an RNA-dependent polymerase that extends the repeated sequences at chromosome ends (Stewart and Weinberg 2006). Most dipterans do not use telomerase; instead, chromosome ends are maintained by repeated transposition of retrotransposons to their chromosome ends (for review see Mason et al. 2008). Dipterans also lack the sequence-specific binding proteins found in organisms that use telomerase (Cenci et al. 2005), and the presence of retrotransposon sequences at chromosome ends is not required for the capping function (Levis 1989; Biessmann et al. 1990; Ahmad and Golic 1998). Despite these differences, in both types of eukaryotes, mutations that cause general telomere dysfunction result in end-to-end chromosome fusions, anaphase bridges, apoptosis, and lethality (Cenci et al. 1997; Nigg 2005). These similarities suggest that telomere loss has the same effect in all eukaryotes, and that all have evolved similar response mechanisms. In fact, the term “telomere,” representing a specialization of the ends of linear chromosomes, was coined by H. J. Muller on the basis of his studies of chromosome biology in Drosophila (Muller 1940).
In mammals, telomerase is not expressed, or is expressed only at low levels in most somatic cells. As a consequence, telomeric repeats grow progressively shorter during rounds of replication and division. Chromosomes lacking telomeric repeated DNA are more frequently involved in end-to-end fusions, and if the cells also lack telomerase they are subject to apoptosis more frequently than their telomerase-positive counterparts (Hemann et al. 2001). This is likely to provide an important check on the proliferation of precancerous cells and is, therefore, of vital interest.
Telomere dysfunction presents the cell with a form of DNA damage that, in some respects, may be mimicked by the double-strand breaks (DSBs) generated by ionizing radiation (IR). Treatment of developing Drosophila with ionizing radiation induces cell cycle arrest and apoptosis. The grp gene, encoding Chk1, is primarily responsible for arrest, with lok, encoding Chk2, contributing to a lesser degree (Xu et al. 2001; Masrouha et al. 2003; Brodsky et al. 2004; De Vries et al. 2005). The apoptotic response is controlled by lok and p53, which encodes the Drosophila homolog of the p53 tumor suppressor (Brodsky et al. 2000a,b, 2004; Ollmann et al. 2000; Xu et al. 2001; Lee et al. 2003; Sogame et al. 2003). To date, no effect of grp on the apoptotic response, nor of p53 on cell cycle arrest, has been observed in Drosophila. Apoptosis is also seen as a response to telomere loss (Ahmad and Golic 1999) and defective telomeres in Drosophila (Queiroz-Machado et al. 2001; Ciapponi et al. 2004, 2006; Oikemus et al. 2004; Silva et al. 2004; Song et al. 2004).
Most experiments that have examined the effects of telomere failure have done so in the context of mutations that cause global telomere dysfunction. The control of the apoptotic response has only been characterized in mutants that cause widespread telomere dysfunction (Song et al. 2004; Musaro et al. 2008). In this circumstance a number of secondary effects can be expected as a result of the end-to-end chromosome fusions that occur (see, for instance, Cenci et al. 1997, 2003; Fanti et al. 1998; Raffa et al. 2005). Moreover, because telomere dysfunction is not limited to a specific developmental stage it is not possible to distinguish the earliest effects from secondary effects.
To specifically examine the cellular responses to loss of a single telomere we use a system that allows us to generate a new chromosome end lacking a telomere during the development of Drosophila melanogaster. The FLP recombinase, whose synthesis is induced by heat shock, causes recombination between inverted FRTs on sister chromatids to generate dicentric and acentric chromosomes (Figure 1; Golic 1994). The acentric chromosome carries the two original telomeres from this arm of the sister chromatids. When the dicentric chromosome breaks in mitosis it delivers a chromosome with a single nontelomeric end to each daughter cell. We refer to this event as telomere loss, to distinguish it from chromosome breaks that may be generated by other means such as irradiation or endonuclease digestion. These other methods will generate two broken ends in a cell, and they may be rapidly rejoined by normal repair machinery (Gong and Golic 2003). The option of rejoining two ends is not available when a cell receives a chromosome with only one broken end. We believe this closely mimics the situation faced by a mammalian cell in which a single telomere becomes critically short and dysfunctional owing to incomplete replication in the absence of telomerase.
Dicentric/acentric chromosome production and segregation. FLP-induced recombination between oppositely oriented FRTs on sister chromatids produces a dicentric chromosome and an acentric chromosome. At anaphase the dicentric chromosome is stretched between the poles and usually breaks. Centromeres are indicated as filled circles, telomeres as filled squares, FRTs as arrows.
This system has a number of advantages. Irradiation makes multiple breaks in random locations, whereas the chromosome arm that experiences telomere loss is determined by the placement of the FRTs. Although the use of rare-cutting endonucleases can provide control over the site of chromosome breakage (Bellaiche et al. 1999; Maggert and Golic 2005), this process still generates two ends. Because the process of FLP-mediated dicentric formation and breakage is extremely efficient it is easily applied in the context of a whole developing animal. Special selective methods are not needed to study cells that have lost a telomere. Our method of dicentric breakage provides a useful model for examining the consequences of single telomere dysfunction.
MATERIALS AND METHODS
The DcX(105) and Dc2(127) dicentric inducible chromosomes were described previously (Golic 1994). The DcY(K2) chromosome was previously called DcYy+ (Ahmad and Golic 1999; also see Ahmad and Golic 1998 for further details of its construction). All others were generated by transposition of a P element carrying inverted FRTs. The locations described in Figure 2 were ascertained by cytology, inverse PCR, or loss of distal markers following FLP induction. The P{70FLP}3F transgene (Golic et al. 1997) is located on the X chromosome, and P{70FLP}4A is located on 3.
Dicentric-inducible chromosomes used in this work. (a–d) The DcY chromosomes: H1 has inverted FRTs inserted near the tip of the long arm; H2 and H3 have inverted FRTs inserted near the tip of the short arm; K2 has inverted FRTs inserted within a duplication of virtually the entire chromosome 4 attached to the long arm of the Y. Chromosomes a and b are based on the standard BSYy+ chromosome (Ashburner et al. 2005). FrTr4B1A is inserted on the short arm of a Y chromosome that also carries y+, although the location of y+ is uncertain. (e–h) X and autosomes with inverted FRT insertions. The location of each insertion is given, along with a schematic indication of the approximate cytological location. The location of Dc3(FrTr1D) is known only approximately by metaphase cytology.
The p53+ transgene was made by amplifying an ∼7.9-kb genome fragment containing the wild-type p53+ gene with 3.2 kb of putative 5′-UTR and 1.2 kb of putative 3′-UTR sequences using the oligonucleotides: CACACACGAATTCCAGATGA and CATCGCTTGGGAAAAGTGCA. The PCR fragment was cloned into the TA cloning vector pGEM-T Easy Vector System I [Promega catalog (cat.) no. A1360]. The fragment was then removed by digestion with the NotI restriction enzyme and cloned into the P element vector pYC1.8{v+} (Fridell and Searles 1991). The p53+ gene was sequenced to ensure that no mutations were introduced by the polymerase. The P element was then transformed by standard methods (Rubin and Spradling 1982).
Flies with lok+ and grp+ transgenes were generously donated by Michael Brodsky.
Immunocytochemistry was carried out as specified in (Laurencon et al. 2003) using the Alexa-Flour 568 from Invitrogen (cat. no. A11036) as the secondary antibody. The anti-phospho-histone H3 antibody was from Upstate (cat. no. 06-570). The cleaved caspase-3 antibody was from Cell Signaling Technologies (cat. no. 9661). Flies were allowed to mate for 5–7 days at which point parents were transferred to a new vial and the larvae were heat shocked in a water bath at 38° for 60 min. Eye and wing imaginal tissue were then dissected out of wandering third instar larvae at ∼12 and 24 hr after heat shock.
Time-lapse microscopy was accomplished using an inverted Olympus IX2-DSU spinning disc confocal and a Hamamatsu Orca-ER digital camera. Embryos were collected and manually dechorionated 0–30 min after being laid. Eggs were then affixed to a coverslip on heptane glue and covered in halocarbon oil.
Neuroblast figures were generated as described (Gatti and Pimpinelli 1983), stained with DAPI, and visualized with a Zeiss Axioplan equipped with an AxioCam HRm and AxioVision software (Golic 1994). A single brain was mounted per slide. Normal and abnormal karyotypes were scored by scanning the entire brain and scoring every metaphase nucleus. Occasional nuclei missing whole chromosomes or sets of chromosomes were considered to be incomplete and were not included in the final tallies.
RESULTS
The mitotic fate of dicentric chromosomes:
In this work, we used several chromosomes carrying inverted repeats of FRTs (Figure 2). The generation of dicentric chromosomes after heat shock induction of FLP is extremely efficient: dicentric chromosomes are produced in ∼90% of somatic cells assayed (Golic 1994), a result that we confirmed (not shown). During anaphase the sister centromeres are pulled apart and the dicentric chromosome is stretched between the spindle poles (Figure 3A). To directly determine the fate of dicentric chromosomes during mitotic divisions we employed time-lapse confocal microscopy. We monitored multiple dicentric-inducible chromosomes in early cellular mitoses from embryos with a maternal contribution of histone-GFP and FLP (Figure 3B). In these experiments >93% of dicentric bridges broke [45/48 generated with DcY(K2), 21/21 bridges generated with DcX(105), and 11/11 generated with Dc2(FrTr1B)]. These results directly confirm the previous conclusion that the predominant fate of dicentric bridges is breakage (McClintock 1941; Ahmad and Golic 1999), with the consequent delivery of a single nontelomeric chromosome end to each daughter cell.
Dicentric bridges and breakage, in vivo. (A) Mitotic divisions in a fixed preparation of the eye imaginal disc [genotype: y w P{70FLP}3F/DcY(K2)]. Chromatin is visualized with anti-phospho-histone H3 antibody shortly after FLP-induced formation of dicentric chromosomes. Anaphase bridges (i, ii, and iv) are easily seen, and one bridge (iii) appears to have broken. (B) Still images from a time-lapse movie of a mitotic division from a cellularized Drosophila embryo that shows breakage of the dicentric bridge. Chromosomes fluoresce because of a His2AvDGFP transgene (Clarkson and Saint 1999). These embryos also carried a dicentric-inducible chromosome and the FLP-expressing transgene [genotype: y w/DcY(K2); P{70FLP}10/+; His2AvDGFP/+].
Control of the apoptotic responses:
In previous work from our lab we observed substantial apoptosis in imaginal discs as early as 8–10 hr after inducing FLP synthesis in flies carrying DcY(K2) (Ahmad and Golic 1999). To investigate the genetic control of this cell death we induced dicentric chromosome formation in wild-type and mutant backgrounds by heat shock induction of FLP and scored for apoptosis in eye (Table 1) and wing (Table 2) imaginal discs 12–14 hr later using an antibody against cleaved caspase-3 (see Figure 4). We observed a strong apoptotic response, which was similar to that which occurs after exposure to 4000 rad of ionizing radiation (Brodsky et al. 2004, and our unpublished results). This response was nearly abolished in p53 mutants, regardless of which dicentric chromosome was tested (Table 1B, Table 2A, ii). Chk2 activates p53 in response to ionizing radiation (Brodsky et al. 2004). In lok mutants the apoptotic response to telomere loss was markedly reduced from wild type (Table 1D, Table 2A, iii). However, in eye discs it is clear that lok mutants still allow more apoptosis than p53 mutants (Table 1, B and D, P = 0.0001), implicating a second pathway in the activation of p53-dependent cell death.
Apoptosis in imaginal discs. Represented here are different severities of apoptosis after dicentric induction. Apoptosis was visualized by indirect immunofluorescence using an antibody against cleaved caspase-3. Dark spots indicate cells undergoing apoptosis in these negative images.
Characterization of apoptosis in the eye imaginal disc at 12–14 hr after induction of dicentric/acentric chromosomes
Characterization of apoptosis in the wing imaginal disc
Chk1 is also activated in the response to ionizing radiation, although it is not thought to be involved in the apoptotic response in Drosophila (Brodsky et al. 2004). We tested for an effect of grp in the response to telomere loss by assaying apoptosis in a lok grp double mutant (Tables 1C and 2A, iv). Cell death was reduced to essentially the same level seen in p53 mutants (Table 1, B and C, P = 0.4056), eliminating the residual apoptosis seen in lok mutants (Table 1, C and D, P = 0.0001) and leading us to conclude that Chk1 does contribute to this p53-dependent apoptotic response, although Chk2 is clearly the primary effector. In mammals, ATR, which is encoded in Drosophila by the mei-41 gene, activates Chk1 following DNA damage (Zhou and Elledge 2000; Shiloh 2003; Helt et al. 2005). Therefore, we tested mei-41 lok double mutants and also found that apoptosis was reduced to a level similar to that observed in p53 mutants (Table 1, D and E, P = 0.0001), confirming that the ATR-Chk1 pathway is involved in the apoptotic response to telomere loss. As with grp, mei-41 mutants show no visible effect on apoptosis unless Chk2 is also removed (Table 1J). In both cases, this may simply reflect our inability to detect a small reduction in apoptosis when the Chk2-dependent pathway is functional.
In mammals, p53, ATR, Chk1, and Chk2 are all capable of mediating cell cycle arrest in response to multiple types of DNA damage (Zhou and Elledge 2000; Shiloh 2003; Artandi and Attardi 2005; Helt et al. 2005; Prieur and Peeper 2008). This suggests the possibility that cells missing a telomere may arrest while attempting to mediate repair of the exposed chromosome end and then succumb to cell competition-induced apoptosis. Hemipterous (hep) is believed to be responsible for mediating this type of apoptosis (Moreno et al. 2002). We tested hep mutants and saw no effect on p53-dependent apoptosis (Table 1K). We also tested a recently isolated suppressor of competition, su(comp)3L-1 (Tyler et al. 2007), and observed no effect on p53-dependent apoptosis (Table 1L). These results indicate that p53-dependent apoptosis is primarily a response to telomere loss and is independent of cell competition or the JNK pathway.
The effect of aneuploidy:
When a dicentric chromosome is generated on the X, or on a major autosome, it is likely to produce significant aneuploidy in the daughter cells owing to the acentric chromosome and asymmetric breakage of the bridge. On the other hand, the dicentric-inducible Y chromosomes are completely dispensable for viability and somatic development, being required only for male fertility. The extensive cell death we observed in these experiments, at 12–14 hr after induction of FLP expression, was virtually identical regardless of whether the dicentric was generated on the dispensable Y chromosome, or on the X, or an autosome (Table 1A and 2A, i). This indicates that the apoptotic response at this time is due to the loss of a single telomere and not a result of chromosome identity or extensive aneuploidy.
In contrast, at later time points we observed additional p53-independent cell death that does appear to be a response to aneuploidy. In p53 or lok mutants, and in lok grp or mei-41 lok double mutants we observed extensive cell death occurring 18–30 hr after FLP induction when dicentrics were generated on the X or an autosome, but not on the Y with DcY(FrTr4B-1A), DcY(H1), DcY(H2), or DcY(H3). Table 2B, ii–v, presents the 24-hr time-point results. Furthermore, if we induced dicentric formation on a Y chromosome capable of generating aneuploidy, DcY(K2), p53-independent cell death was also seen, although to a lesser degree than with the X or autosomal dicentrics. The inverted FRTs on DcY(K2) are embedded in a duplication of chromosome 4 on the tip of the long arm. It is, therefore, capable of generating aneuploidy after dicentric breakage, either as a consequence of an asymmetric break that distributes unequal numbers of the FRT-proximal chromosome 4 genes to the two daughter cells or by accumulation of acentric fragments carrying the FRT-distal genes (see next section). Considered as a whole, our data indicate the delayed p53-independent cell death is specific for aneuploidy.
Previous work that characterized the cellular response to IR also revealed two temporally separate apoptotic pathways. The first is p53 and Chk2 dependent and is activated as early as 4 hr after irradiation (Brodsky et al. 2004). This appears to correspond to the primary telomere-loss response that we see following dicentric formation on any chromosome. Our data indicate that this response is specific for the detection of the broken chromosome end. The delayed apoptotic response to IR does not require p53, Chk2 (Wichmann et al. 2006), Mei-41, or Chk1 (our unpublished data), and almost certainly corresponds to the aneuploid-specific cell death we observe at a similar time point after dicentric induction.
Cells that escape apoptosis:
The observation of a highly proficient apoptotic response to telomere loss does not necessarily indicate that every cell with a chromosome that experiences telomere loss will undergo death, or whether such death will follow immediately, or after another round of division, or more. To assess the ability of cells to escape apoptosis following telomere loss we analyzed karyotypes of larval neuroblasts at several time points after the induction of FLP synthesis (Figure 5, Table 3). At 24 hr after 70FLP induction only ∼10–20% of cells showed normal karyotypes. In flies with intact apoptotic response pathways the percentage of cells with normal karyotypes increased rapidly so that by 96 hr after heat shock ∼80–90% of cells showed normal karyotypes. However, in flies lacking p53 or the upstream activators Chk2 and Mei-41, cells with normal karyotypes did not exceed 40%, even 96 hr after 70FLP induction.
Frequency of the wild-type karyotype in larval brains after dicentric induction. The y-axis shows the percentage of cells with a wild-type karyotype among all metaphase spreads in the entire larval brain at different time points after dicentric induction as denoted on the x-axis. Usually 3–4 brains were counted with an average of 64 metaphases per brain. Genotypes and complete data are given in Table 3.
Karyotype aberrations induced by dicentric chromosome formation
The analysis of abnormal karyotypes allowed us to draw some conclusions about the mitotic history of a cell that experienced dicentric formation. A cell in the first metaphase immediately following dicentric formation can be identified by its characteristic dicentric/acentric karyotype (Figure 6A; Golic 1994). Certain other karyotypes are indicative of cells that reach subsequent metaphases. For instance, a centric chromosome fragment that is visibly longer or shorter than the original dicentric chromosome must be derived by the breakage of a dicentric bridge in a previous division (Figure 6B). The presence of acentric chromatids in numbers other than one also indicates cells that have reached a subsequent metaphase (Figure 6, B–D, and I). In other words, a cell bearing two or more acentric chromatids, or an altered dicentric-inducible chromosome with no acentric chromosome, must have reached at least the first metaphase subsequent to the anaphase in which the original dicentric chromosome broke. Most cells with abnormal karyotypes had two acentric chromosome segments, but a significant fraction had more than two (Table 3). At all time points, and in all genotypes, cells with two or more acentric chromatids outnumbered cells with none (among cells showing evidence of dicentric formation). We observed many cells with multiple and varying numbers of acentric chromosomes, including one cell with 54 acentric chromatids (Figure 6I). Even though a cell with this karyotype is grossly aneuploid, because each acentric chromosome segment carries two copies of a large portion of the X chromosome, it must have reached at least the fifth metaphase subsequent to the initial division with a dicentric and an acentric chromosome. Similar aberrant karyotypes were observed following dicentric induction on other chromosomes. Figure 6C shows a cell with 14 acentrics recovered 24 hr after dicentric induction on the Dc2(FrTr1B) chromosome, which must have reached at least the third metaphase subsequent to the initial division with a dicentric and acentric chromosome. The karyotypes that indicate a cell divided multiple times after dicentric formation were more frequent in flies that had an impaired apoptotic response, but they also occurred in flies with intact apoptotic pathways (Table 3).
Larval neuroblasts with altered karyotypes after dicentric formation. The genotype of each cell is indicated, along with the time after 70FLP induction. Acentric chromatids of expected size are indicated with arrowheads. (A–D) Karyotypes of cells with Dc2(FrTr1B) [genotype: y w P{70FLP}3F/y w; Dc2(FrTr1B)/+]. (A) A cell at the first metaphase showing dicentric (Dc2) and acentric of chromosome 2 (Ac). (B) A cell showing two acentric chromatids and a short centric fragment (Fr2) with apparently unfused ends. (C) A cell with 14 acentric fragments, including some that differ in size (yellow arrow). (D) A cell showing four circular acentric chromatids. (E and F) Karyotypes of mei-41 chk2 cells with Dc3(FrTr1D) [genotype: mei-4129D/Y; lokp6; P{70FLP}4A/Dc3(FrTr1D)]. (E) An octaploid cell for all chromosomes except the fourth, carrying a stable Y;3 translocation that was successfully replicated three times (yellow arrows). (F) A tetraploid cell with an end–end fusion between two chromosome 3′s that was successfully replicated (yellow arrow). (G–I) Karyotypes isolated from cells with DcX(105). (G and H) Telophase nuclei exhibiting lagging acentric chromatids [genotype: y w P{70FLP}3F/DcX(105); p535A-1-4]. (I) A nucleus with 54 acentric chromatids of three different sizes (white, yellow, and red arrows) [genotype: y w P{70FLP}3F/DcX(105)].
Our estimate of mitotic history is conservative, and is based on the assumption that all acentric chromosomes replicate and segregate to the same daughter cell at each division. Furthermore, we assumed that such a daughter cell also carried a long centric fragment of an asymmetric breakage event. This long fragment will still carry inverted FRTs, and, if FLP perdures, another round of FLP-mediated recombination can again generate acentric and dicentric chromosomes. It seems unlikely that all of these assumptions will be satisfied. For instance, Figure 6G shows six condensed acentric chromosomes located between two interphase nuclei with decondensed chromosomes, apparently indicating a failure of acentrics to segregate after having successfully segregated for at least two rounds of division. We do not believe this is simply an example where the remainder of a complete metaphase genome is located elsewhere or lost. There were no other metaphase spreads within one optical field in any direction (63× objective, F.N. 25). Furthermore, we observed other instances of decondensing chromosomes at the poles with condensed acentric chromosomes in the middle (Figure 6H). We conclude that acentric chromosome segments do not always faithfully segregate to a single daughter cell. In such cases, our calculations will underestimate the actual number of divisions that a cell experienced after dicentric formation.
We also saw chromosome structures that indicate cells have experienced bridge–breakage–fusion cycles. We observed several examples of nuclei with ring acentric chromosomes (Figure 6D). Such structures almost certainly arose by three steps: (1) asymmetric breakage of a dicentric chromosome; (2) replication of the long centric chromosome and fusion of sister-chromatid broken ends; (3) FLP-mediated recombination between inverted FRTs on sister chromatids to form a free circle (Figure 7). We observed many other mitotic figures where sister chromatid ends were fused together, similarly indicating that a cell with a broken chromosome had passed through G1 and into S, and then the replicated broken ends fused. Such a cell is expected to experience another round of bridge and breakage.
A mechanism for circular acentric chromosome formation. Symbols are as in Figure 1.
We sometimes observed acentric chromosomes of different sizes in the same nucleus (Figure 6, C and I). These were seen in cells with an intact apoptotic response and in cells where this response was impaired. These may have arisen by further rounds of FLP-mediated recombination in a cell carrying the long product of asymmetric breakage. The size of the acentric piece would then be determined by the site of breakage in the previous anaphase. Different sizes of acentric chromosomes could also be generated by fusion between a broken end and a de novo double-strand break on another chromosome, leaving the remainder of that other chromosome as a new fragment that is subject to rearrangement and mis-segregation. Several examples of apparent heterologous fusions were seen, with one having replicated multiple times (Figure 6E), supporting the idea that acentric chromosomes might arise in such a manner. Heterologous fusions were never seen in the absence of FLP induction.
Tetraploid and polyploid cells were also seen (Figure 6, E and F), although at a much lower rate. The occurrence of tetraploidy increased two- to eightfold in mutant cells when compared to wild-type cells (Table 3). For instance, with DcX(105) we observed 3 tetraploid nuclei in 12 wild-type brains, but 22 tetraploid nuclei in 11 p53 brains. Similar result were obtained with Dc2(FrTr1B) and Dc3(FrTr1D). Tetraploidy was never seen in the absence of FLP induction.
DISCUSSION
Our results show that loss of a single telomere from only one chromosome leads to levels of apoptosis that are similar to those seen after 4000 rad of ionizing radiation. This apoptosis is independent of chromosome identity, occurs irrespective of whether aneuploidy is produced, and is not the result of cell competition, indicating that it is a specific response to telomere loss.
Pathways to cell death:
Telomere-loss-induced cell death is mediated by the activation of p53 via Chk2 and Chk1. Since the nontelomeric chromosome end is produced as a result of anaphase bridge breakage, the first opportunity that a cell has to detect the problem is during telophase or G1 of the cell cycle. Our finding that the majority of cell death is dependent on Chk2 is consistent with robust detection in G1. Chk2 can operate throughout the cell cycle, in contrast to Chk1 whose primary role appears to be limited to S and G2 phases (Zhou and Elledge 2000; Shiloh 2003; De Vries et al. 2005). We hypothesize that most telomere loss events are detected rapidly in G1 with Chk2-dependent apoptosis as the ultimate outcome. Cells that escape G1 detection may proceed into S phase, where detection of the nontelomeric end provokes Chk1-dependent apoptosis. Cells that escape both checkpoints may proceed to the next mitosis.
It has been shown that, following irradiation, apoptosis can also occur by a p53-independent mechanism (Wichmann et al. 2006). Our results show that this p53-independent death occurs only when aneuploidy is produced. Thus, there can be a third check on the proliferation of cells with a broken chromosome. This aneuploidy-dependent mechanism is delayed by many hours relative to the p53-dependent apoptosis, likely indicating that it is an effect of altered gene dosage and requires gene expression for the effects of altered stoichiometry to be felt. This death may result from the same mechanism that kills slow-growing Minute cells in a Minute//Minute+ mosaic tissue (Moreno et al. 2002).
Escaping cell death:
It is of exceeding interest to learn how cells escape any or all of these three checkpoint mechanisms. We found that cells that had divided one or more times after receiving a chromosome with a nontelomeric end were not rare. In the most extreme case we observed a cell that appeared to have divided at least four times after receiving a broken chromosome and was preparing to divide a fifth time.
One method for cells to escape checkpoint detection would be to add a new telomere to the broken end. An efficient mechanism to repair a broken end by this route has been documented in the male germline (Ahmad and Golic 1998). Whether such a mechanism exists in the soma is still an open question, but it is not the only mechanism that allows a cell to escape apoptosis. Cells that survive to S phase and replicate the broken end may temporarily solve the problem by fusing those ends, leading to a bridge–breakage–fusion cycle (McClintock 1941). Our karyotype analyses provided evidence for such events. Such cells should be subject to detection and elimination in subsequent cell cycles, although this did not always happen.
It is quite possible that p53-dependent apoptosis is simply not 100% efficient. Higher than normal levels of p53+ expression might be needed to achieve 100% apoptosis in response to telomere loss. However, overexpression of p53+ causes apoptosis in otherwise normal cells (Ollman et al. 2000; Lee et al. 2003; Colombani et al. 2006), suggesting that p53 is finely tuned at a level that allows normal cells to survive, but that is insufficient to cause the death of all cells that experienced telomere loss. Even if the p53-mediated response is <100% efficient it is clearly critical for the elimination of cells with a nontelomeric chromosome end. The p53-independent response produces lower levels of apoptosis and is relatively ineffective at eliminating cells with abnormal karyotypes.
Aneuploidy may underlie another mechanism of escape from apoptosis. Our observation of cells with extreme aneuploidy suggests the possibility that the altered gene dosages present in these cells may cause serious disturbances in the normal functioning of cell cycle and DNA damage checkpoint systems and allow them to escape death despite the continued presence of a nontelomeric end. The evidence that aneuploidy may affect normal cell cycle controls has recently been discussed (Torres et al. 2008). We were surprised to discover that many cells retained acentric chromosome segments and, in some cases, they accumulated to large numbers. However, evidence for the segregtion of acentric chromosomes has been obtained in yeast (Kaye et al. 2004), insects (Platero et al. 1999; Lafountain et al. 2001) and human cells (Portnoi et al. 1999; Kanda et al. 2001), although there is no single clear mechanism.
Finally, it is possible that some cells may adapt, as yeast cells can (Sandell and Zakian 1993; Lee et al. 1998) and resume growth despite the persistence of a broken end.
Genome instability:
The loss of a single telomere is capable of inducing persistent genomic instability in cells that survive. We observed examples of repeated rearrangement of the chromosome that had originally lost a telomere. There was clear evidence for fusion events involving other chromosomes, expanding the reach of instability to include more of the genome than just the original dicentric chromosome. We also observed instances of tetraploidy and higher levels of ploidy. All aberrant karyotypes were more frequent in cells with a defective apoptotic response. Our results add to previous findings from yeast and mammalian cell lines showing that a single dysfunctional telomere is sufficient to induce instability (Sandell and Zakian 1993; Lee et al. 1998; Hackett and Greider 2003; Sabatier et al. 2005).
Much evidence suggests that genome instability is an early event in carcinogenesis (Nishizaki et al. 1997; Kolquist et al. 1998; Al-Mulla et al. 1999; Buerger et al. 1999; Shpitz et al. 1999; Artandi and Depinho 2000; Artandi et al. 2000; Romanov et al. 2001). Chromosome rearrangements, such as those we observed, may promote transformation by mutation or altered regulation of genes at or near breakpoints and by altering gene dosage through large-scale duplication and deletion. If any of these events cripple the apoptotic checkpoints a higher percentage of cells with aberrant genomes will survive, providing greater opportunity for carcinogenic events to arise and propagate. Thus, loss of a single telomere may be a precipitating event in carcinogenesis.
Modeling telomere loss:
Telomere-defective mutants have been identified in many model organisms and have also been produced in cultured cell systems (Nigg 2005; Stewart and Weinberg 2006). In general, these mutations cause genome instability through end-to-end chromosome fusions. Prominent effects of these mutations in animal systems are shortened lifespan or lethality and an altered susceptibility to cancer. But in spite of the extensive interest in telomeres and their function there have been relatively few attempts to model the loss of a single telomere. However, this is probably the most frequent circumstance for a cell to encounter a defective telomere. As somatic cells divide without expression of telomerase, telomeric repeat sequences progressively shorten owing to incomplete replication. Since chromosomes vary in the length of their telomeric repeat DNA, one telomere is likely to reach a critically short length before the rest. The first cellular effects of telomere shortening are a response to persistent short telomeres in the cell (Hemann et al. 2001). Learning how cells respond to this event is likely to be critical in understanding the earliest stages of aging and cancer.
In yeast and cultured mammalian cells, the genomic consequence of single telomere loss has been examined by placement of a negatively selectable gene near the end of a chromosome (Sandell and Zakian 1993; Sabatier et al. 2005). Loss of the chromosome end can be stimulated by using HO or I-SceI endonuclease to introduce a double-strand break. After selection, examinations of the surviving cells have revealed that they have either lost the chromosome or carry healed chromosomes, where telomeric repeats have been added to the end of the truncated chromosome. Many chromosomes also show evidence of a period of instability from bridge–breakage–fusion events. In other cases the chromosome may have acquired a telomeric end by translocation, and in the process, transferred instability to another chromosome.
There have been even fewer studies that have examined the effect of single telomere loss in the context of a developing organism, where a mixture of normal cells and cells that have lost a telomere may interact in the normal organismal environment. This allows the assessment of cellular dynamics in a realistic situation that may not be easily reproduced by unicellular organisms or by cells in culture. McClintock used a set of well-placed genetic markers to follow the fate of cells in the developing corn plant after loss of a single telomere. When a pollen grain delivered a chromosome with a broken end to the female gametophyte she observed continuing bridge–breakage–fusion cycles in the endosperm, but healing of the broken end in the zygote (McClintock 1941). Developmental regulation of telomerase expression appears to be general in plants (Fitzgerald et al. 1996), perhaps accounting for this difference. In the work reported here, we followed cell fate in developing Drosophila with a combination of cell-death measurements and karyotype analyses and found evidence for the occurrence of bridge–breakage–fusion cycles in the soma. In previous work we observed efficient healing of a chromosome broken in the male germline and transmitted to offspring (Ahmad and Golic 1998). The analogy between McClintock's findings in plants and ours in an animal is difficult to ignore. There are parallels in the expression of telomerase in humans: functional telomerase is expressed in the germline but not in most somatic cells (Shay and Wright 2001). The degree to which the healing behavior of somatic and germline cells is conserved is unknown. It is certainly possible, even likely, that there are undiscovered selective steps that intervene between the generation of a broken end in a germline cell and the recovery of viable offspring. Similarly, we do not rule out the possibility that healing may also occur as one response to detection of a nontelomeric end in the soma. One advantage of studying the response to telomere loss in a whole animal is that such questions are open to examination.
Acknowledgments
We thank Michael Brodsky, Nicholas Baker, Robert Saint, William Theurkauf, and Yikang Rong for their generosity with fly lines; William Sullivan and Uyen Tram for assistance with confocal microscopy; and Mary Golic, Heng Xie, Rebeccah Kurzhals, Carlos Diaz-Castillo, and Sienna Berry for technical assistance. This work was supported by the National Institutes of Health grant RO1 GM65604 and the National Institutes of Health Genetics Training Grant T32-GM007464.
Footnotes
Communicating editor: T. C. Kaufman
- Received July 4, 2008.
- Accepted September 30, 2008.
- Copyright © 2008 by the Genetics Society of America