Main

Telomeres are tracts of DNA consisting of G-rich tandemly repeated sequences (5′-TTAGGG-3′ in vertebrates) found at both ends of every chromosome. Telomeres undergo gradual shortening during the proliferation of normal somatic cells, thereby setting an upper limit on the number of times that cells can divide. One of the hallmarks of cancer is the presence of a cell population with unlimited growth potential, which is enabled by a telomere lengthening mechanism (TLM) capable of completely counteracting normal telomere shortening. A substantial number of cancers, including some that are biologically aggressive and currently difficult to treat, use an HR-mediated DNA-copying mechanism, referred to as ALT, to prevent telomere attrition1,2. Here we review recent advances in understanding of the nature of the ALT mechanism and how it is repressed in normal cells.

Mechanism of ALT

It has long been known that telomeres can be lengthened by telomerase, which synthesizes telomeric DNA by reverse transcription of an RNA template3. ALT is defined as any nontelomerase mechanism of telomere lengthening1. Most of what is known about ALT in higher eukaryotes has been obtained from study of human telomerase-negative immortalized cell lines and cancers (for brevity, we refer to these as 'ALT cells' and 'ALT cancers', respectively). The model first proposed was that ALT involves synthesis of new telomeric DNA on the basis of a DNA template4, and this model is now supported by several lines of evidence. The model in which ALT involves copying of telomeric DNA templates is based on observations that a DNA tag inserted into one telomere becomes copied to other nonhomologous telomeres in human ALT cells and in normal mouse cells5,6,7 and may also become duplicated in its original location in human ALT cells8. The simplest interpretation of these observations is that a telomere can use itself or a telomere on a sister chromatid or other chromosome as a copy template.

This mechanism is essentially the same as the process subsequently proposed to repair breaks at chromosome termini, referred to as break-induced replication, in which only one end of the break invades a template (reviewed in ref. 9). It has been demonstrated in the yeast Kluyveromyces lactis that circles of extrachromosomal DNA can be used as a template for telomere extension10. Human ALT cells have abundant extrachromosomal telomeric DNA, in linear, circular and complex forms (reviewed in ref. 11). The ALT mechanism may therefore potentially include copying of any telomeric DNA sequences, although to date only chromosomal telomeric DNA has been demonstrated to be a template in human cells.

The hypothesis that ALT involves HR is supported by evidence that genes encoding HR proteins (including Rad50, Rad51, Rad52 and Sgs1) are essential for telomerase-negative yeast survivors, and HR proteins (including MRE11, RAD50 and NBS1 (which form the MRN complex), SMC5 and SMC6 (which form a heterodimer) and MMS21, FEN1, MUS81, FANCA and FANCD2) are also essential for telomere-length maintenance in human ALT cells (reviewed in ref. 11). Circumstantial evidence has been provided by the observation that in ALT cells many HR proteins (including RAD51, RAD52, RPA, NBS1, SLX4, BLM, MRN, BRCA1 and BRIT1) are present, along with telomeric DNA and telomere-binding proteins, in a subset of promyelocytic leukemia (PML) bodies referred to as ALT-associated PML bodies (APBs) (refs. 12,13,14,15 and citations in ref. 11). Therefore, the main steps involved in ALT most probably include strand invasion of the template molecule and formation of an HR intermediate structure (Holliday junction (HJ)), copying of the template, dissolution of the recombination intermediates and possibly synthesis of the complementary strand (Fig. 1).

Figure 1: Four putative steps in the ALT mechanism.
figure 1

(1) Strand invasion of the template molecule and formation of an HR intermediate structure. (2) Copying. (3) Dissolution of the HR intermediate. (4) Possible filling in of the complementary strand (red dashed line). The complementary strands of the DNA duplex are depicted as red and blue lines.

Full size image

The single-stranded (ss) G-rich overhang at the telomere terminus has the potential to invade homologous DNA, but this first step in the ALT mechanism is normally prevented by coating of telomeres with shelterin proteins, including POT1, which binds to telomeric ssDNA (reviewed in ref. 16). Initiation of the ALT process may therefore require a switch from loading of the overhang with POT1 to loading with RAD51, a protein that promotes strand invasion and annealing to the complementary strand17. What triggers this switch is not clearly understood, but an intermediate event is likely to be the replacement of POT1 with replication protein A (RPA). Telomeric repeat–containing RNA (TERRA) and heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1) have been shown to be involved in the switch between binding of RPA and POT1 to the telomere during the cell cycle18. Displacement of RPA from ssDNA filaments by RAD51 requires the assistance of mediator proteins (reviewed in ref. 19).

The polymerase responsible for the second step, template-directed synthesis of telomeric DNA, is not known but could be polymerase δ or ζ, given the putative role of these enzymes in HR20,21. In this regard, PCNA, an accessory protein for polymerase δ, has been found in APBs22, and the polymerase δ subunits Pol31 and Pol32 are both required for break-induced replication in Saccharomyces cerevisiae23.

The third step is the processing of the intermediate products of HR, such as HJs, and this step must occur before chromosome segregation. HJ-processing complexes include GEN1, SLX4–SLX1 and MUS81–EME1, which are responsible for HJ resolution, and the BLM–TOP3A–RMI1–RMI2 (BTR) HJ-dissolution complex24,25,26. Competition between HJ-dissolution and HJ-resolution complexes is likely to be involved in the processing of telomeric HR intermediates during ALT. Both SLX4 and BLM interact directly with TRF2 (refs. 12,13,27,28), thus suggesting that these two proteins and their complexes may compete for binding to telomeres. SLX4 assembles multiple endonucleases at telomeres, thereby regulating processing of HR intermediates. It has recently been shown that BLM depletion results in increased telomere sister-chromatid exchange events, whereas SLX4 depletion decreases the number of these events28, and that SLX4 negatively regulates telomere length in ALT cells13. Therefore, BLM and SLX4 may compete for the processing of HR intermediates at the telomere, such that BLM promotes ALT activity through telomere synthesis and HJ dissolution, and SLX4 inhibits ALT activity through HJ resolution in the absence of telomere synthesis.

It is currently unknown whether the postulated fourth step in the ALT process—filling in of the underhanging strand—occurs, or whether the elongated strand simply acts as a template for semiconservative DNA replication in the next S phase of the cell cycle. If this step does occur, the relevant polymerases remain to be identified.

Adjacent and distant templates for ALT activity

The template for HR-mediated telomeric copying may be a sister chromatid or another region of the same telomere8 (Fig. 2). It has been proposed that telomeres may copy themselves either by looping out or by looping back on themselves. Looping out of the G-rich strand may increase the likelihood of G-quadruplex formation.

Figure 2: Adjacent and distant copy templates in ALT.
figure 2

(a–c) Use of a nearby copy template in the sister-chromatid telomere (a) or in the same telomere by looping out (b) or by t-loop formation (c) is depicted. (d) A distant telomere may be used as a copy template after Hop2–Mnd1–dependent long-range movement35.

Full size image

When the gene encoding the TRF2 shelterin protein is deleted experimentally, telomeres exhibit a DNA-damage response (DDR) and become more mobile within the cell nuclei29. Consistently with the observation that many telomeres within ALT cells spontaneously associate with DDR foci30, a subset of ALT telomeres display substantial directional movements and the ability to dynamically associate with each other and with APBs31. APBs, which are suspected to be hubs of telomere recombination in ALT cells32,33, are induced after treatment with DNA-damaging agents34.

Recently, in ALT cells a specialized homology-directed searching mechanism has been identified that drives long-range rapid and directional movement and subsequent clustering of telomeres35. Expression of a TRF1-FokI fusion protein has been used to create telomere-specific DNA double-strand breaks (DSBs), and direct measurements of telomere trajectories have identified a subset of highly mobile telomeres that move across micrometer-range distances and connect with relatively constrained telomeres and APBs35. Telomere movement is dependent on the Hop2–Mnd1 heterodimer35, which stabilizes the RAD51–ssDNA presynaptic filament or stimulates the HR pairing reaction36. At present, it is not clear what proportion of ALT activity requires long range-movement versus involvement of immediately adjacent telomeric sequences (Fig. 2).

The role of telomeric chromatin in repressing ALT

Telomeric DNA is tightly packaged into nucleosomes with a linker-DNA length approximately 40 bp shorter than that in bulk chromatin37. Nucleosomes are dynamic structures that disassemble and reassemble to allow replication-fork progression. Nucleosomes are particularly mobile at telomeric DNA sequences, and mobility can be induced by the shelterin protein TRF1, which alters nucleosome assembly by binding to both the linker DNA between nucleosomes and nucleosomal binding sites38 A direct comparison of isogenic telomerase-positive and ALT cell lines has identified a lower nucleosome density associated with ALT activity39, thus implicating telomeric nucleosome disorganization in ALT.

Nucleosome mobility is also enhanced by ATP-dependent chromatin-remodeling complexes and histone chaperones. Co-depletion of the ASF1 paralogs ASF1a and ASF1b, which are histone chaperones that interact with histone regulator A (HIRA) and the chromatin assembly factor 1 (CAF1) complex, results in a striking induction of ALT phenotypic markers in both mortal and telomerase-positive cells7. Intriguingly, this phenomenon has been observed only in cells with long telomeres, and it results in concomitant suppression of the human telomerase reverse transcriptase transcript and thus inhibition of telomerase activity. The mechanism of ALT induction after ASF1 depletion is thought to involve perturbation of histone recycling at replication forks and to be dependent on ATR protein kinase signaling in response to RPA loading7. Although ASF1 mutations have not been identified in ALT cell lines, these experiments demonstrate the importance of normal chromatin architecture in repressing ALT in normal cells.

Telomeric chromatin is normally enriched in repressive histone modifications, including heterochromatic histone H3 K9 trimethylation, histone H4 K20 trimethylation, HP1 binding and histone hypoacetylation (refs. 40,41; reviewed in ref. 42). Histone methyltransferase–knockout studies in mice have suggested a role for decreased telomeric heterochromatin in telomere elongation and increased telomeric recombination43. Although the potential role of increased telomerase activity has not been ruled out, these observations have led to the notion that normal telomeric heterochromatin formation represses ALT and conversely that perturbations may provide a recombination-permissive environment that facilitates ALT. However, no global differences in either histone acetylation or methylation between human telomerase-positive and ALT cells have yet been identified15.

How ALT becomes derepressed in cancer

Genetic changes.

A striking correlation has been observed between ALT activity in various human cancers and loss of the ATP-dependent helicase ATRX or its binding partner, the H3.3-specific histone chaperone DAXX, both of which are constituents of PML bodies44,45,46. This correlation has been shown to be robust in a set of 41 pancreatic neuroendocrine tumors, of which 25 (61%) were ALT positive; all of the ALT tumors had mutations in the ATRX or DAXX genes or had abnormal expression of their encoded proteins. In liposarcomas and cancers of the central nervous system, mutations in ATRX (or less commonly DAXX) also correlate with the presence of ALT activity44,47,48. A survey of ALT cell lines has identified abnormal ATRX expression in 86% of ALT cell lines45. Further investigation is required to determine whether some ALT tumors retain normal ATRX and DAXX expression, but it is clear that loss of normal ATRX–DAXX function is highly correlated with ALT activity.

Cellular immortalization studies have identified loss of ATRX as an early event during ALT activation after cellular crisis, and reexpression of ATRX suppresses ALT activity49, thus indicating that ATRX is an ALT repressor. However, knockdown or knockout of ATRX is insufficient to switch telomerase-positive cell lines to telomere maintenance by the ALT pathway or to directly activate ALT in mortal cell strains7,45,49. Although it has been shown that ATRX–DAXX is involved in deposition of histone H3.3 at telomeres of mouse embryonic stem cells50, that ATRX binds to telomeric repeats and G-quadruplex structures51, that ATRX facilitates replication through aberrant DNA secondary structures52 and that ATRX is involved in telomere cohesion53, it is not yet clear which, if any, of these functions are required for repressing ALT in normal cells. The identification of ALT repressors such as ATRX–DAXX supports speculation that ALT may be an ancient TLM that has persisted in eukaryotes (Box 1).

Changes in telomeric nucleoprotein structure.

ALT telomeres display an elevated telomere-specific DDR, which is partly independent of telomere length and can be suppressed by exogenous expression of TRF2 (ref. 30). This elevated DDR appears to be due to interrelated changes in telomeric DNA sequence and binding proteins. The telomeres of human ALT cell lines have a substantially increased content of variant (noncanonical) telomeric repeats54,55, a phenomenon that may be explained as follows. The distal ends of human telomeres are composed predominantly of tandem arrays of the TTAGGG repeat unit, whereas the proximal regions contain many variant repeats56,57. In ALT cells, HR-mediated copying has been shown to occur at sites throughout the telomeric-repeat array58 and to result in stochastic amplification of canonical and variant repeats and ultimately the spread of variant repeats throughout the telomeres54,55.

These altered telomere sequences in ALT cells result in altered telomeric protein binding, in part because of the loss of shelterin-binding sites. The shelterin complex maintains telomere integrity by inhibiting DDR, chromosome fusions and HR (reviewed in ref. 16), and the elevated frequency of DDR foci at ALT telomeres has been attributed to decreased shelterin binding30. Altered DNA sequences may also create new protein-binding sites40,54,59. For example, TCAGGG variant repeats create a high-affinity binding site for a group of nuclear hormone receptors, including TR2, TR4, COUP-TF1, COUP-TF2 and EAR2, which are abundant at ALT telomeres40,54,59. TR4 and COUP-TF2 recruit the zinc-finger protein ZNF827, which binds via its N-terminal RRK motif to the NuRD nucleosome-remodeling and histone-deacetylation complex15, and thereby has an extensive series of ALT-promoting effects. NuRD–ZNF827 causes telomeric histone hypoacetylation, which has been hypothesized to partly counteract other changes that promote telomeric decompaction15. In addition, NuRD–ZNF827 facilitates telomere-telomere tethering and the recruitment of HR proteins, including the DDR protein BRIT1 and the HR-repair protein BRCA1 (ref. 15).

ALT cells have recently been found to have telomeric sequences enriched in TCAGGG repeats associated with nuclear-receptor binding, located at multiple sites throughout the genome59. The telomere-genome interactions identified in that study59 may potentially be explained by the binding of nuclear receptors to the variant repeats and the consequent recruitment of NuRD–ZNF827, which can tether disparate sites (Fig. 3). It has been shown experimentally that interstitial insertion of telomeric DNA can destabilize the genome60, so the dispersion of telomeric sequences found in ALT cells may account, at least in part, for the ongoing genomic instability in ALT cells that is manifested by karyotypic rearrangements and micronuclei45,61.

Figure 3: Multiple roles of the NuRD–ZNF827 complex at ALT telomeres.
figure 3

Binding of nuclear receptors to variant repeats in ALT telomeres, and perhaps elsewhere in the genome, results in recruitment of ZNF827 and the NuRD chromatin-remodeling and histone-deacetylation complex. This displaces shelterin binding, tethers different telomeric sequences together and recruits proteins involved in HR15.

Full size image

Role of telomeric replication stress in ALT

Normal telomeres are intrinsically challenging to the DNA-replication machinery: aphidicolin-induced replication stress has revealed telomeres to be fragile sites62. This may reflect the propensity for G-rich repetitive DNA to form secondary structures such as G quadruplexes, DNA-RNA hybrids, D loops, t loops and HJs, all of which hinder replication-fork progression. Stalled forks must be repaired to maintain genetic stability; accordingly, numerous proteins that process DNA intermediate structures have been identified at telomeres. These include the helicases BLM, WRN, RecQL4 and RTEL1 (refs. 63,64), the exonuclease Apollo (reviewed in ref. 16) and the scaffold protein SLX4, which recruits many structure-specific endonucleases12,13,26. These proteins dismantle unusual DNA structures, thus allowing unimpeded replication-fork progression65.

Binding of telomeric DNA to the shelterin complex appears to aid in telomere replication, and TRF1 has an essential role in preventing the appearance of telomere fragile sites and the formation of stalled replication forks62. In the event of persistent replication defects or replication-fork collapse, various cellular pathways cooperate to repair the DNA damage and resume cell-cycle progression or to trigger programmed cell death. Key pathways involved in telomere DDR include ATM and ATR signaling, and this signaling is partly suppressed by members of the shelterin complex. Specifically, TRF2 represses the response by ATM to DSBs, whereas POT1 represses ATR signaling at ssDNA, which can arise at DSBs or at stalled replication forks66.

ALT telomeres may cause an even greater level of replication stress, especially when shelterin binding is depleted by the presence of variant sequences and thus of competing binding proteins. As noted above, a substantial proportion of the telomeres in ALT cells are associated with a DDR focus. Interestingly, a recent study has found that ALT cancer cells are hypersensitive to ATR inhibitors, thus resulting in chromosome fragmentation and apoptosis; hence, there may be increased ATR signaling in ALT cells67. Consequently, replication stress may paradoxically increase ALT activity by stimulating HR.

Conclusions and perspective

The past few years have seen substantial advances in understanding of the molecular details of how ALT becomes derepressed in cancer, notably the loss of function of the ATRX–DAXX complex, the increased proportion of noncanonical repeats within telomeres and the consequent telomeric recruitment of different binding proteins (especially nuclear receptors, some of which recruit NuRD–ZNF827) and changes in telomeric chromatin. The dependence of ALT cells on the DDR protein ATR and the molecular details of the long-range movements of damaged telomeres have been uncovered.

Many questions about ALT remain to be answered, including the following. Why is ALT preferentially activated in specific cancer subsets? What is the molecular basis for repression of ALT by ATRX–DAXX? And what are the genetic changes that result in derepression of ALT when ATRX–DAXX is not mutated? A more detailed understanding of the ALT mechanism and how it is controlled may ultimately be exploited to develop effective anticancer therapeutics.