Telomere flip-flop: an unfolding passage to senescence

EMBO Rep (2011) advance online publication. doi:; DOI: 10.1038/embor.2011.227

Telomeres can be thought of as a double-edged sword that both helps and hinders human health. On the one hand, telomeres provide an important tumour suppressor mechanism by limiting how many times a human cell can divide. On the other hand, very short telomeres can promote tumour progression by allowing rampant genome instability. In this issue of EMBO reports, Reddel and colleagues provide new insight into how telomeres can behave in such a Dr Jeckyll and Mr Hyde type manner by demonstrating that cell division is halted when about five telomeres become dysfunctional but the dysfunction is not necessarily the result of permanent damage. Instead, the telomeres seem to experience a transient defect that can be reversed even in the absence of telomerase. Only when telomeres become much shorter do they experience the lethal damage or ‘uncapping’ that results in chromosome fusions and scrambled chromosomes.

Human primary cells have a limited proliferative lifespan because they lack telomerase and hence undergo telomere shortening with each cell division (Verdun & Karlseder, 2007). Eventually the shortened telomeres are sensed as DNA damage and trigger a cell cycle checkpoint that halts cell division, thereby initiating replicative senescence. However, if p53 or pRb (retinoblastoma) are inactivated, the DNA damage checkpoint is circumvented and cell division proceeds. The telomeres continue to erode until eventually they become too short to protect the DNA termini from the repair proteins responsible for non-homologous end joining (NHEJ). The cells then enter a period called crisis during which the chromosomes undergo end-to-end fusion and enter a breakage–bridge fusion cycle that leads to further genome rearrangements (Murnane, 2011).

The DNA damage signalling and repair processes that occur after full telomere uncapping have been studied extensively because it is relatively easy to achieve acute loss of telomere protection by removal of a telomere-binding protein (Takai et al, 2003). This exposure of the DNA termini results in activation of ATM (ataxia telangiectasia mutated) and/or ATR (ataxia telangiectasia and Rad3 related) signalling followed by DNA end processing and NHEJ (Stewart et al, 2011). It has been more difficult to study the events leading to replicative senescence and the physical nature of telomere dysfunction because only a subset of telomeres are affected and direct visualization of these telomeres has been technically challenging. Nonetheless, it has become apparent that with increased population doubling a growing number of telomeres accumulate γH2AX and 53BP1, characteristics of DNA damage (d'Adda di Fagagna et al, 2003; Zou et al, 2004). Cell cycle arrest is ultimately triggered by a pathway involving ATM, p53 and pRb signalling (Gire et al, 2004; Herbig et al, 2004; Takai et al, 2003; Zou et al, 2004). As cells enter senescence, the accumulation of telomeric γH2AX correlates with an increase in the number of chromosomes with very short telomeres, implicating short telomeres as the cause of the DNA damage response (Zou et al, 2004). However, most chromosomes retain sufficient telomeric DNA to be visible by telomere FISH (fluorescence in situ hybridization) and the few chromosomes that lack telomere signals do not undergo end-to-end fusions, indicating that the DNA terminus cannot be completely exposed to the DNA repair machinery. Thus, the physical defect leading to γH2AX accumulation and damage signalling in senescent cells has remained a mystery. Moreover, as cells approach senescence they continue to divide despite having accumulated multiple γH2AX-marked telomeres. This finding suggests that dysfunctional telomeres signal less effectively than DNA double-stranded breaks, and a certain threshold number must accumulate before the signal is sufficient to trigger cell cycle arrest (Zou et al, 2004).

The current manuscript by Reddel and co-workers (Kaul et al, 2011) addresses both the number of dysfunctional telomeres needed to induce senescence and the nature of the telomere dysfunction. The work builds on a key advance made by the Reddel lab (Cesare et al, 2009) that allows simultaneous visualization of telomeric DNA and telomere proteins or repair factors on metaphase chromosomes. The technique involves swelling cells before a cytospin so the metaphase cells break open as they are deposited on a microscope slide. The spread chromosomes can then be analysed for both telomeric DNA content by FISH and protein localization by indirect immunofluorescence. Previously, such co-localization was limited to whole cells or nuclei because the spreading technique routinely used for karyotype analysis requires fixatives that denature antigenic determinants. With whole-cell analysis, γH2AX foci that co-localize with telomeric DNA (termed TIFs or telomere dysfunction-induced foci) can be counted, but it is not possible to determine whether the remaining foci mark dysfunctional telomeres that lack telomeric DNA or merely sites of DNA damage at non-telomeric loci. Moreover, because it is hard to resolve individual telomeres within a cell, it is difficult to determine their length by quantitative FISH. With the new ‘meta-TIF’ analysis, one can count the total number of dysfunctional telomeres per metaphase and quantify their telomeric DNA content.

Kaul et al have used the meta-TIF analysis to examine the accumulation of dysfunctional telomeres with increased population doubling in a variety of human primary cells. They show that regardless of tissue origin, young cells exhibit a significant number of meta-TIFs. This number increases with increased population doubling up to a maximum, at which point the culture enters senescence. Intriguingly, most of the meta-TIFs are present on only one sister chromatid (chromatid-type meta-TIF) rather than on both sisters (chromosome-type meta-TIF; see Fig 1A). As senescent cells mostly reside in G₀/G₁ of the cell cycle, the authors use the number of chromosome- and chromatid-type meta-TIFs to calculate the average number of G₁ TIFs present when cells become senescent. This comes out at about five per cell. However, if p53 is inactivated and the population lifespan increases, so do the number of metaphase and G₁ TIFs. Thus, human primary cells have a p53-dependent threshold of DNA damage signalling corresponding to an average of five dysfunctional telomeres. p53-positive cells can keep dividing until the threshold is reached, whereas p53-deficient cells continue to divide well beyond this point.

Figure 1.

Outcomes of telomere dysfunction. (A) Fate of chromatid-type meta-TIFs with permanent compared with transient telomere dysfunction. Telomeres are yellow, TIFs are red. (B) Progressive telomere uncapping with increasing population doubling. The fully protected telomeres in young cells prevent DNA damage signalling and NHEJ. Partly unprotected telomeres in senescent cells prevent NHEJ but not damage signalling. During crisis, fully unprotected telomeres trigger damage signalling and NHEJ. NHEJ, non-homologous end joining.

The work also provides interesting insight into the nature and fate of TIFs during the cell cycle. The authors confirm previous results suggesting that dysfunctional telomeres might aggregate in interphase cells (Zou et al, 2004). Analysis of large interphase TIFs revealed that some contain several telomeric FISH signals and the overall intensity of the signal is greater than that expected for a single telomere. The authors also show that some meta-TIFs are repaired after cells enter G₁, and that meta-TIFs are brighter and easier to visualize than interphase TIFs. These results explain why TIFs have not previously been observed in young cells and why interphase cells generally seem to have fewer TIFs or dysfunctional telomeres than metaphase cells.

One key observation concerns the fate of the chromatid-type meta-TIFs with increasing population doubling. Because chromatid-type meta-TIFs occur on only one sister, they are thought to arise from replicative or post-replicative damage. If the damage underlying meta-TIF formation is permanent, a chromatid-type meta-TIF should be converted to a chromosome-type meta-TIF in the next cell cycle when the parental chromosome is replicated to give two sisters with damaged telomeres (Fig 1A), but this was not observed. Instead, the fraction of chromosome-type meta-TIFs increased only slightly and remained small even in senescent cells. This finding indicates that the TIFs do not represent permanent damage and that they can be repaired in the next cell cycle.

Further analysis of the meta-TIFs revealed more striking characteristics. First, 60–90% of the meta-TIFs from senescent cells retained telomeric DNA that was visible by FISH. Second, although meta-TIFs were often observed on the shorter of the two sister telomeres, in some cases the meta-TIF was present on the sister with the longer telomere. Third, telomeres with meta-TIFs retained the telomere proteins TRF2 and RAP1, implying that they were still capped by the protective shelterin complex. These observations suggest that telomeres with meta-TIFs retain many features of a functional telomere. However, the presence of γH2AX clearly indicates that these telomeres are somehow deficient. What elicits the DNA damage response is unknown, but a probable answer is that the telomeres have become less able to fold into the protective chromatin structure that normally forms after DNA replication and that is needed to sequester the chromosome terminus from the DNA damage response machinery (Fig 1B). This structure might involve T-loop formation, but it could also involve some other type of heterochromatic marks. Telomere shortening might increase the frequency of misfolding because of the decreased abundance of telomere proteins that promote or stabilize the folded structure. However, the occurrence of misfolding seems to be stochastic; so although a telomere might experience a problem in one cell cycle, it might fold correctly in the following cell cycle.

Overall, the essential contribution of this work lies in its compelling support for the ‘partial uncapping’ or three-state model for telomere end protection. This model has been proposed in various forms to explain why telomere shortening triggers senescence but telomere fusions only accumulate in cells that circumvent senescence and continue to divide (Cesare et al, 2009; Verdun & Karlseder, 2007). As now demonstrated by Kaul et al (2011), telomeres from senescent cells retain sufficient telomeric DNA and telomere proteins to confer protection from NHEJ, but not enough to prevent DNA damage signalling. Consequently, telomere fusions would only occur after bypass of senescence led to further telomere shortening and complete loss of the protective telomeric DNA–protein complex. This stepwise removal of telomere protection provides an elegant mechanism to halt cell division before telomere shortening endangers genome stability and explains how the ends of human chromosomes can function both as tumour suppressors and tumour promoters.