p53- and ATM-Dependent Apoptosis Induced by Telomeres Lacking TRF2

See allHide authors and affiliations

Science  26 Feb 1999:
Vol. 283, Issue 5406, pp. 1321-1325
DOI: 10.1126/science.283.5406.1321


Although broken chromosomes can induce apoptosis, natural chromosome ends (telomeres) do not trigger this response. It is shown that this suppression of apoptosis involves the telomeric-repeat binding factor 2 (TRF2). Inhibition of TRF2 resulted in apoptosis in a subset of mammalian cell types. The response was mediated by p53 and the ATM (ataxia telangiectasia mutated) kinase, consistent with activation of a DNA damage checkpoint. Apoptosis was not due to rupture of dicentric chromosomes formed by end-to-end fusion, indicating that telomeres lacking TRF2 directly signal apoptosis, possibly because they resemble damaged DNA. Thus, in some cells, telomere shortening may signal cell death rather than senescence.

Mammalian telomeres consist of several kilobases of tandem TTAGGG repeats bound by the related telomere-specific proteins, TRF1 and TRF2 (1). TRF1 regulates telomere length (2) and TRF2 maintains telomere integrity (3). Inhibition of TRF2 results in loss of the G-strand overhangs from telomere termini and induces covalent fusion of chromosome ends (3, 4).

To investigate the cellular consequences of telomere malfunction, we used adenoviral vectors to overexpress intact and truncated versions of TRF1 and TRF2 (Fig. 1) (5). These vectors encoded full-length TRF1 (AdTRF1); a dominant negative version of TRF1 lacking the Myb DNA binding domain (AdTRF1ΔM); full-length TRF2 (AdTRF2); an NH2-terminal deletion of TRF2 lacking the TRF2-specific basic domain (AdTRF2ΔB); and a dominant negative version of TRF2 lacking both the basic domain and the Myb domain (AdTRF2ΔBΔM). Transformed and primary cells were infected efficiently, resulting in overexpression of the different TRF proteins at approximately the same levels (6).

Figure 1

Adenovirally expressed TRF1 and TRF2 proteins. Expression of the indicated TRF proteins was monitored by immunoblotting, and their localization was determined by IF (6). Inhibition of TRF1 and TRF2 was determined by gel-shift assay (9) and IF (3), respectively. Dicentric formation was deduced from the presence of anaphase bridges in infected cells.

Indirect immunofluorescence analysis (2, 3, 7) indicated that TRFs retaining the Myb domain localized to telomeres, as deduced from their characteristic punctate staining pattern in interphase nuclei (Fig. 1) (6). In contrast, TRFs lacking the Myb domain did not accumulate at telomeres but displayed a dispersed nuclear staining pattern. Expression of AdTRF1ΔM reduced the TRF1-specific TTAGGG repeat binding activity in cell extracts (8, 9), implying that this allele has the expected dominant interfering effect on the DNA binding activity of endogenous TRF1. The dominant negative effect of AdTRF2ΔBΔM was documented based on displacement of the endogenous TRF2 protein from the telomeres and its ability to induce chromosome end fusions (Fig. 1 and Table 1) (3).

Table 1

Cell type dependence of AdTRF2ΔBΔM-induced apoptosis.

View this table:

AdTRF2ΔBΔM resulted in a rapid induction of apoptosis in infected HeLa cells, as assessed by TUNEL labeling, Annexin-V staining, and the appearance of cells with a sub-G1 DNA content in fluorescence activated cell sorting (FACS) analysis (10) (Fig. 2, A and B). HeLa cells transfected with an unrelated TRF2ΔBΔM expression construct lacking adenoviral sequences had the same response, demonstrating that apoptosis did not require adenoviral gene products. None of the other TRF alleles affected cell viability or growth in the short term, although a minor G2 accumulation phenotype occurred in HeLa cells expressing full-length TRF1 and TRF2 (Fig. 2B) (11). The absence of apoptosis with AdTRF2 and AdTRF2ΔBsuggested that AdTRF2ΔBΔM induced apoptosis by reducing the amount of TRF2 bound to telomeres.

Figure 2

Induction of apoptosis by AdTRF2ΔBΔM. (A) TUNEL labeling (10) of HeLa and IMR-90 cells 72 hours after infection with the indicated viruses. AdTRF2ΔBΔM induced apoptosis in HeLa cells 48 hours after infection. (B) FACS analysis (10) of HeLa and IMR-90 cells 72 hours after infection with the indicated adenoviruses. Y-axis: cell numbers; X-axis: relative DNA content, based on staining with propidium iodide. (C) Immunoblot analysis (15) of p53, Bax, and cyclin D1 (as a loading control) in HeLa and IMR-90 cells infected with the indicated adenoviruses.

In contrast to HeLa cells, the HDF cell strain IMR-90 and the fibrosarcoma cell line HT-1080 did not undergo apoptosis after inhibition of TRF2 (Fig. 2A) (3). To explore the cell-type dependence of the apoptotic response, we infected several immortalized human cell lines and primary human cell strains with AdTRF2ΔBΔM. Apoptosis was observed in two HeLa subclones, the p16-deficient mammary adenocarcinoma cell line MCF7, immortalized B cells (see below), and primary CD4+ T cells (Table 1). Cells that failed to undergo apoptosis in response to AdTRF2ΔBΔM either lacked an apoptotic response to DNA damage (12) or were deficient for p53 function (13). Although HeLa cells are partially compromised for p53 function because they express human papilloma virus-16 E6 tumor antigen, which targets p53 for degradation, they do contain a fully functional p53 pathway (14). To investigate this pathway, we examined protein levels of p53 and its downstream target Bax in HeLa cells expressing the five different TRF alleles (15). p53 levels rose only in response to AdTRF2ΔBΔM and there was a concomitant increase in the expression of Bax (Fig. 2C), indicating that p53 is activated as a transcription factor. Induction of p53 and Bax was also observed in cells not undergoing apoptosis.

The role of p53 was confirmed using mouse embryo fibroblasts (MEFs) from genetically altered mice. The dimerization domains of human TRF2 and mouse Trf2 are nearly identical (1), suggesting that the mouse protein would be inhibited by AdTRF2. Indeed, wild-type MEFs infected with AdTRF2ΔBΔM lost the punctate nuclear pattern typical of Trf2 engaged on telomeres and showed chromosome end fusions (Table 2), validating the use of AdTRF2ΔBΔM in murine cells (16, 17). Wild-type MEFs and MEFs from p21 (18), INK4a/ARF (19), or RB (20) knockout mice showed a strong apoptotic response to AdTRF2ΔBΔM (∼18% apoptosis) (Fig. 3A and Table 2), while uninfected controls and cells infected with AdTRF2 or AdTRF2ΔB showed background levels of apoptosis [2 to 5%; (Table 2) (17)]. In contrast, MEFs derived from p53–/– mice (21) did not show this apoptotic response (Fig. 3A and Table 2). The AdTRF2ΔBΔM protein was clearly active as a dominant negative allele in the p53–/– MEFs, since the cells displayed a high rate of anaphase bridge formation. These data establish a requirement for p53 function in the telomere-mediated apoptotic pathway.

Figure 3

AdTRF2ΔBΔM-induced apoptosis requires the ATM/p53 pathway but not progression through S phase. (A) Absence of AdTRF2ΔBΔM-induced apoptosis in MEFs lacking p53. Wild-type and p53–/– cells were analyzed by TUNEL labeling (10) 72 hours after infection with AdTRF2ΔBΔM (16, 17). (B) A-T B cells did not show AdTRF2ΔBΔM-induced apoptosis. Human B cell lines from a normal donor (GM130C) and an A-T patient (GM01526E) expressing AdTRF2ΔBΔM were analyzed for apoptosis by TUNEL labeling (10) at 24 hours after infection. (C) Immunoblot (15) of p53, p21, and cyclin D1 levels in B cells derived from normal and A-T donors 20 hours after infection with the indicated adenoviruses or γ-irradiation. (D) Progression through S phase is not required for the induction of apoptosis. Top, schematic of the experiment. Bottom, detection of BrdU uptake and TUNEL labeling in T cells infected with the indicated adenoviruses (24).

Table 2

Requirement of p53 and ATM for AdTRF2ΔBΔM-induced apoptosis.

View this table:

p53 can be phosphorylated by the ATM kinase, the product of the gene that is mutated in patients with ataxia-telangiectasia (A-T) (22). ATM is thought to function upstream of p53 in a DNA damage response pathway because A-T cells show diminished induction of p53 and reduced ability to arrest in G1 after ionizing radiation (23). To explore the possibility that ATM also functions in the telomere-directed apoptotic response documented here, we analyzed the effect of AdTRF2ΔBΔM on the viability of human B cell lines from A-T patients and normal donors. Normal B cell lines rapidly underwent apoptosis, whereas B cells from A-T patients did not show this response (Fig. 3B and Table 2). In addition, paralleling their response to γ-irradiation, B cells from A-T patients showed no evidence of p53 induction or up-regulation of the p53 target p21 after infection with AdTRF2ΔBΔM (Fig. 3C). Thus, similar to the signaling pathway activated by double-stranded breaks, the induction of apoptosis by telomere malfunction requires ATM, most likely as an upstream activator of p53.

Two types of events might be responsible for the activation of the ATM/p53-dependent apoptotic pathway. Because loss of TRF2 function induces covalent fusion of telomeres (3, 4), mitotic rupture of dicentric chromosomes will generate DNA breaks that can activate a DNA damage response. However, a signal could also emanate directly from the chromosome ends, if telomeres lacking TRF2 resemble damaged DNA. To distinguish between these possibilities, we determined whether progression through mitosis is required for the AdTRF2ΔBΔM-induced apoptosis. Primary human peripheral T cells were harvested in a quiescent state, stimulated to proliferate in the presence of bromodeoxyuridine (BrdU), and infected with AdTRF2ΔBΔM and AdTRF2 prior to entry into S phase (24). Cells were simultaneously examined for BrdU incorporation and apoptosis 24 hours later (Fig. 3D). Although adenovirus infection inhibited T cell proliferation slightly, a substantial fraction of the cells progressed through S phase and incorporated BrdU. As expected, AdTRF2ΔBΔM, but not AdTRF2, induced apoptosis in primary T cells. While some cells showed dual labeling for both BrdU and TUNEL, a substantial proportion (∼80%) of the apoptotic cells in the AdTRF2ΔBΔM-infected cultures did not contain detectable amounts of BrdU (Fig. 3D). Thus, a subset of the AdTRF2ΔBΔM- infected cells entered apoptosis before progressing through S-phase, and hence before undergoing mitosis. This interpretation is corroborated by the drop in BrdU-positive cells in AdTRF2ΔBΔM-infected T cells (Fig. 3D) (24). Based on these data, we conclude that in primary human T cells, inhibition of TRF2 function can induce apoptosis independent of the damage resulting from broken dicentric chromosomes.

Collectively, these data reveal a previously unrecognized function of mammalian telomeres: their ability to prevent the induction of apoptosis by chromosome ends. We considered the possibility that inhibition of TRF2 might result in sudden loss of telomeric DNA, leaving uncapped chromosome ends that could activate a DNA damage checkpoint. However, HeLa and IMR-90 cells infected with AdTRF2ΔBΔM showed no loss of TTAGGG repeat DNA detectable by standard genomic blotting. Thus, binding of TRF2 to the TTAGGG repeats is apparently required to “mask” chromosome ends from a response pathway that can lead to cell death. The exact nature of the apoptotic signal from the unmasked telomeres remains to be established. The involvement of ATM suggests that inappropriately exposed telomeric DNA might resemble a double-stranded break, which is the predominant initiating signal for ATM-mediated p53 activation [reviewed in (25)]. Cells expressing TRF2ΔBΔM rapidly lose the 3′ overhang of telomeric TTAGGG repeats that protrudes from human telomeres (3), so conceivably, this unique feature of telomeres may be a criterion by which random double-stranded breaks are distinguished from natural chromosome ends. However, we cannot exclude the existence of an ATM/p53-dependent telomeric checkpoint that directly senses the altered status of the telomeric complex in TRF2ΔBΔM-expressing cells.

Compromised telomere function leads to deregulation of telomere length, fusions of chromosome ends, and senescence (2, 3,26). Our data indicate that at least in some mammalian cell types, telomere malfunction can also induce apoptosis, a suggestion consistent with the increased rate of apoptosis in mice with critically shortened telomeres (27). Thus, disintegration of the telomeric complex as a consequence of telomere erosion in the human soma may cause apoptosis of some cells (such as T cells) and senescence of others (such as fibroblasts). Since telomeres also shorten during tumorigenesis (28), induction of the apoptotic pathway described here may provide an additional selection for malignant cells that are deficient in p53 function.

  • * These authors contributed equally to this work.

  • Present address: Fox Chase Cancer Center, 7701 Burholme Avenue, Philadelphia, PA 19111, USA.

  • Present address: Chiron Corporation, 4560 Horton Street, Mailstop 4-3, Emeryville, CA 94608–2916, USA.

  • § To whom correspondence should be addressed. E-mail: delange{at}


View Abstract

Navigate This Article