Recombination-dependent maintenance of telomeres, first discovered in budding yeast, has revealed an alternative pathway for telomere maintenance that does not require the enzyme telomerase. Experiments conducted in two budding yeasts, S. cerevisiae and K. lactis, have shown recombination can replenish terminal G-rich telomeric tracts that would otherwise shorten in the absence of telomerase, as well as disperse and amplify sub-telomeric repeat elements. Investigation of the genetic requirements for this process have revealed that at least two different recombination pathways, defined by RAD50 and RAD51, can promote telomere maintenance. Although critically short telomeres are very recombinogenic, recombination among telomeres that have only partially shortened in the absence of telomerase can also contribute to telomerase-independent survival. These observations provide new insights into the mechanism(s) by which recombination can restore telomere function in yeast, and suggest future experiments for the investigation of potentially similar pathways in human cells.
Linear chromosomes must have a mechanism to ensure that their termini are fully duplicated when the chromosome itself is replicated. For most eukaryotes, the ends of nuclear chromosomes are maintained by the reverse transcriptase telomerase. Telomerase adds species-specific telomeric repeats onto chromosome termini, thereby counterbalancing sequence loss that would ensue as a result of semi-conservative DNA replication by the conventional DNA replication machinery (for reviews, see Greider, 1996; McEachern et al., 2000). This not only ensures complete replication of the genome, but the telomeric repeats that are synthesized by telomerase provide a binding site for the host of telomere-specific proteins that prevent recognition of chromosome termini as double strand breaks (for reviews, see Lundblad, 2000; Blackburn, 2001).
However, increasing interest has been focused on a recombination-dependent mechanism for replenishing telomeric repeats. This pathway, first discovered in budding yeast, is receiving attention due to the possibility that a potentially similar pathway operating in human cells, referred to as ALT (for Alternative Telomere Lengthening), might contribute to neoplastic transformation. This review discusses the investigations in budding yeast that have helped elucidate this recombination-dependent process, followed by a discussion of the possible mechanisms that have been proposed as a result of the yeast studies. A brief summary of the ALT pathway, and a discussion about whether ALT proceeds through a similar recombination-dependent pathway, concludes this review. The reader is also referred to other excellent reviews on this topic (Biessmann and Mason, 1997; Kass-Eisler and Greider, 2000; Reddel et al., 2001; Kraus et al., 2001).
Discovery of a telomerase-independent pathway for telomere maintenance
Several observations during the 1980′s provided early clues that recombination could potentially maintain telomeric repeats at chromosome ends. Linear plasmids that lacked a functional telomere, but contained homology to sub-telomeric regions, could acquire a functional chromosome terminus in a homology-dependent process (Dunn et al., 1984). Rescue of these linear plasmids was proposed to involve strand invasion of the sub-telomeric region of a chromosome by the terminus of the linear plasmid, followed by replicative copying of the donor onto the end of the plasmid (Figure 1a). The concept that a genetic recombination intermediate could be converted into a replication fork had already been proposed as a mechanism for replication of the ends of the linear bacteriophage T4 genome (Luder and Mosig, 1982). Formosa and Alberts (1986) provided biochemical support for the contribution of recombination proteins to the formation of a replication fork, and further suggested that homology-initiated DNA replication might also apply to the telomere end replication problem. Similar recombination-based models for telomere maintenance had also been elaborated by Bernards et al. (1983) and Walmsley et al. (1984).
Direct demonstration that recombination could maintain the natural termini of linear chromosomes, however, awaited the analysis of telomerase-defective yeast strains. The discovery of the S. cerevisiae EST1 gene (Lundblad and Szostak, 1989), later shown to encode a subunit of the telomerase holoenzyme (Hughes et al., 2000), allowed examination of the consequences of a telomere replication defect in an easily experimentally manipulated organism. A haploid yeast strain deleted for the est1-Δ gene exhibited gradual telomere shortening and an accompanying progressive decline in viability, dubbed yeast cellular senescence. This established an experimental paradigm for the relationship between telomere replication and limits on replicative capacity, which was recapitulated in human cells several years ago (Bodnar et al., 1998). In addition, the rate of loss of a non-essential reporter chromosome increased, in parallel with the appearance of est1− senescence (Lundblad and Szostak, 1989). Since chromosome loss is the most probable outcome of a DNA double strand break (Kramer and Haber, 1993; Sandell and Zakian, 1993), this suggested that cell death in an est1-Δ cell occurs when telomeres shorten sufficiently that the ends of chromosomes resemble double strand breaks. A more recent study has expanded the potential lethal events that occur in an est1-Δ cell (Hackett et al., 2001). Their analysis indicates that cell death may result from loss of essential genes (as a result of extensive terminal deletions) and/or chromosome transmission failures resulting from end-to-end fusions that trigger a breakage-fusion-bridge cycle.
Although the majority of yeast cells die in the absence of telomerase, Lundblad and Blackburn (1993) observed that a small subpopulation could escape the lethal consequences of a telomerase defect. The rare survivors recovered from an est1-Δ strain not only regained the ability to grow but also displayed dramatic changes at their chromosomal termini, due to global amplification of both telomeric and sub-telomeric repeat sequences. These extensive rearrangements arose as a result of recombination, as the appearance of survivors was blocked if the est1-Δ strain was also defective for RAD52, which is responsible for the majority of homologous recombination events in yeast. Survivors could be grouped into two general classes based on the pattern of amplification of telomeric and/or sub-telomeric sequences. Although survivors were recovered as healthy growing clones arising from a culture of mostly inviable cells, continued single colony propagation of both classes of survivors revealed variable growth patterns: sub-clones could be isolated in which a senescence phenotype re-appeared, although survivors readily reappeared again from these senescing sub-clones. Highly dynamic changes in telomere structure, with sequences rapidly lost or acquired at individual telomeres of survivor strains, was also observed, providing a possible explanation for the growth variability. Finally, neither est1− strains early in senescence nor subsequently isolated survivors displayed a genome-wide hyper-recombination phenotype. However, this analysis did not exclude the possibility that in a cell that lacked telomerase, unreplicated telomeres could themselves become hyper-recombinogenic, a point that will be re-addressed later in this review.
Although all survivors recovered from an est1-Δ strain shared common features, such as RAD52-dependence and highly dynamic changes at their telomeres, they could be grouped into two classes based on notable differences in their telomeric structure. One class of survivors was distinguished by extensive amplification of sub-telomeric repeats (called Y′ elements). In most wild type strains of S. cerevisiae, approximately 2/3 of the telomeres contain tandem arrays of one to four Y′ elements, separated by short stretches of ∼50 to 100 bp of G-rich telomeric repeats (Figure 2a). In a telomerase-proficient strain, exchanges of Y′ elements between telomeres had been shown to occur at low frequency, resulting in the occasional dispersal of these repeats among chromosome ends (Horowitz and Haber, 1985). However, in one class of survivors, Y′ copy number increased dramatically (200-fold in some survivor strains), forming extended tandem arrays at many telomeres (Figure 2a). In addition, telomeres that did not have any Y′ elements in the pre-survivor est1− strain acquired Y′ elements. Sub-telomeric repeat amplification was also accompanied by a correspondingly substantial increase in the total amount of telomeric DNA (due to the short stretches of telomeric tracts bracketing each Y′ element), with up to 4% of the genome composed of telomeric repeat DNA in some survivors. However, despite the overall increase in telomeric DNA, the terminal G-rich repeat tract remained short (and did not appear to undergo further shortening) in this class of survivors.
This analysis also identified a second class of est1− survivors, with a distinctively different telomeric restriction pattern: survivors in this group were characterized by changes in the multiple sizes of telomeric restriction bands, relative to the parental strain. Survivors in this second class did not exhibit extensive sub-telomeric repeat amplification, although many, if not all, telomeres appeared to acquire at least partial Y′ sequences. The full characterization of the nature of the telomeric changes occurring in this class of survivors came from subsequent work by Teng and Zakian (1999), discussed in more detail in the next section.
Based on the above observations, Lundblad and Blackburn (1993) proposed that, in the absence of telomerase, recombination could maintain G-rich telomeric repeats and restore telomere function. We further suggested that conversion to a telomerase-independent survivor was a multi-step process that involved multiple rounds of recombination, occurring during the growth of an est1− culture, with each beneficial recombination event providing a slight selective advantage. This model implied that there was a gradient of telomere function, such that any reduction in telomere length could reduce the replicative capacity of the est1− culture, and any recombination event that restored function to even a single telomere could contribute to improved growth. Subsequent work showed that a strain with a partially defective telomerase, resulting in stably short telomeres but no senescence, had a severe growth disadvantage relative to a telomerase-proficient strain, as assessed by liquid competition experiments (Morris and Lundblad, 1997). Alternative models, which propose that critically short telomeres are primarily responsible for triggering senescence and serving as substrates for recombination, are considered later in this review.
For the class of survivors that displayed Y′ amplification, Y′ elements were proposed to be acquired by either a nonreciprocal translocation event (Figure 3a), or by integration of extrachromosomal Y′ circles into a critically short telomere (Figure 3b). The latter model was based on the demonstration that Y′ elements, which contained a weak origin of replication (Chan and Tye, 1983), could be propagated as extrachromosomal plasmids (Horowitz and Haber, 1985). Expansion of sub-telomeric Y′ elements to multiple telomeres in this class of est1− survivors would have two consequences. First, this would increase the extent of homology between telomeres, thereby enhancing subsequent recombination events and continuing to drive the process. In addition, the amplified arrays of Y′-associated telomeric repeats would increase the amount of G-rich telomeric repeat DNA present at each chromosome, which provides a reservoir of G-rich telomeric repeat DNA for recombination onto the extreme terminus of chromosomes.
Recombination between terminal G-rich telomeric repeats, without Y′ amplification, also appeared to be capable of maintaining the terminal G-rich tract (Figure 3c). This was inferred from the observation that even survivors that exhibited extensive amplification of Y′ elements still possessed individual telomeres composed only of G-rich repeats, with no newly acquired Y′ element. These G-rich tracts therefore appeared to be maintained by telomere–telomere recombination. Thus, the key feature of the overall model was that the acquisition of G-rich telomeric repeats through recombination was responsible for restoring telomere function and viability.
Subsequent work from several laboratories has shown that the appearance of survivors is not specific to yeast strains that are defective for EST1 function. Yeast strains lacking any telomerase subunit (Est1p, Est2p, Est3p or TLC1), or defective in the telomerase recruitment function of Cdc13p, all produce survivors in a recombination-dependent manner that is indistinguishable from that observed for est1− survivors (Singer and Gottschling, 1994; Lendvay et al., 1996; Teng and Zakian, 1999).
Recombination-mediated telomere maintenance can elongate terminal telomeric G-rich repeats
Analysis of the consequences of a telomerase defect in a related budding yeast, Kluveromyces lactis, revealed additional insights regarding how recombination could maintain telomeres. In K. lactis, loss of telomerase function had the same consequences as in S. cerevisiae: gradual attrition of telomeric repeats, accompanied by a senescence phenotype (McEachern and Blackburn, 1995). Survivors could also be recovered from senescent cultures of telomerase-deficient K. lactis strains, through a mechanism that was again dependent on RAD52 function (McEachern and Blackburn, 1996). However, analysis of telomere structure in these survivors uncovered some notable differences. Unlike S. cerevisiae, K. lactis telomeres do not have tandem arrays of sub-telomeric repeats interspersed with G-rich telomeric repeat sequences (note that, unlike telomeric repeats, sub-telomeric repeat elements show little conservation from species to species). The absence of sub-telomeric repeats meant that a class of survivors analogous to the S. cerevisiae survivors that exhibited extensive Y′ amplification was not recovered in K. lactis. Instead, the terminal G-rich telomeric tract was greatly elongated in telomerase-defective K. lactis survivors (Figure 2b). Different telomeres in post-senescent survivors exhibited variable lengths, often with tract lengths longer than telomeres of a telomerase-proficient strain. However, these elongated telomeres still exhibited the same gradual shortening observed during the initial senescence of a newly generated telomerase-defective strain, suggesting that the recombination mechanism that was replenishing telomeric repeats was not as efficient as that provided by telomerase. As with S. cerevisiae survivors, K. lactis survivors also displayed an unstable growth phenotype, manifested as secondary episodes of transient senescence.
These observations led McEachern and Blackburn (1996) to present a model for telomere repeat array lengthening (Figure 1b) similar to that proposed by Dunn et al. (1984) for the healing of linear plasmids (Figure 1a). They proposed that the telomere shortening that occurs in the absence of telomerase results in loss of the essential ‘capping’ function of a telomere, so that chromosome ends are now perceived as double strand breaks. Processing of these exposed termini by DNA repair enzymes could produce a 3′ single stranded overhang that could invade another chromosomal telomere. Extension of this invading strand, followed by second strand synthesis, would result in a non-reciprocal translocation, with sequences from the donor chromosome transferred to the ‘uncapped’ chromosome. These gene conversion events could be the result of invasion of a 3′ strand that still retains homology with a telomeric repeat tract on another chromosome; staggered alignment of the invading and donor strands would generate a telomeric tract longer than that observed in telomerase-plus cells.
Although the primary structural change observed in K. lactis survivors was extension of the telomeric tract array, gene conversion events in sub-telomeric regions of K. lactis survivors were also observed. Since 11 of the 12 K. lactis chromosome termini share homology in the sub-telomeric region, this indicates that strand invasions could initiate well into sub-telomeric regions. However, McEachern and Blackburn pointed out that the majority of the recombination events that they observed must depend on retention of at least a few telomeric repeats at the time that chromosome termini undergo recombination-mediated elongation.
Through a careful molecular analysis, Teng and Zakian (1999) showed that the second class of S. cerevisiae survivors identified by Lundblad and Blackburn (1993) exhibited many of the properties displayed by survivors recovered from K. lactis. Telomeres in this second class of S. cerevisiae survivors, which they re-named type II survivors (while survivors with extensive Y′ amplification were labeled type I survivors), were shown to terminate with very long, heterogeneous tracts of G-rich telomeric repeats (Figure 2a). As had been previously observed in K. lactis, these extended telomeric tracts were not stably maintained: they displayed the slow attrition of a telomerase-defective strain, as well as more abrupt changes that were reminiscent of previous observations showing that elongated telomeric tracts could undergo rapid truncation (Li and Lustig, 1996; Bucholc et al., 2001). Thus, although type I and type II survivors both retain terminal G-rich repeat tracts, terminal sequences are greatly elongated in type II survivors. Notably, the short tracts that are retained on the telomeres of type I survivors are apparently more stable than the continually shortening type II telomeric tracts (although it cannot be ruled out that additional shortening of a type I telomere might be lethal and thus would not contribute to the signal on a telomeric Southern blot).
Teng and Zakian also studied in detail the growth properties of the two classes of survivors. In their strain background, all type II survivors exhibited a stable growth phenotype, and when propagated in culture, type II survivors also had a selective advantage over type I survivors. The growth disadvantage of type I survivors might be due – at least in part – to the burden of replicating a genome increased by up to 10%, due to the extensive amplification of Y′ and telomeric repeat sequences. However, type I survivors also showed a high propensity for conversion to type II survivors, suggesting that telomeres in the absence of telomerase may be substrates for several different types of recombination events. The eventual outcome may be dictated by a balance between the susceptibility of an under-replicated telomere to different types of recombination events, versus the selective advantage of different recombinant telomeric structures. In other words, telomeres in budding yeast may be more recombinogenic for the types of events that result in type I survivors, but the selective growth advantages of type II survivors allows this class of survivors to dominate. This may have implications for the analysis of potential intermediates for similar telomere maintenance pathways in human cells.
Do critically short telomeres promote the recombination-dependent pathway?
The original model by Lundblad and Blackburn proposed a stochastic process of recombination events at telomeres, with selection pressure for viability ensuring that there would be a cumulative acquisition of telomeric sequences. A variant of this model is that telomeres become more recombinogenic in the absence of telomerase. A further extension of the model is that recombination occurs in a burst, triggered by critically short telomeres, rather than a succession of cumulative exchanges. Recent studies from several different laboratories suggest that short telomeres are in fact recombinogenic, even before critically shortening or senescence is observed (McEachern and Iyer, 2001; Rizki and Lundblad, 2001). However, additional observations indicate critically short telomeres may also be preferred substrates for at least one type of recombination (Teng et al., 2000). Thus, both variations of this model are likely to contribute to telomerase-independent telomere maintenance.
An initial clue that recombination-mediated exchanges were important for cell survival even early in the propagation of a telomerase-deficient strain came from observations of the consequence of loss of RAD52 function. A rad52− mutation not only blocked the appearance of survivors, but in addition, an est1-Δ rad52-Δ strain also displayed a greatly accelerated senescence phenotype (Lundblad and Blackburn, 1993). Whereas an est1-Δ strain at the 25 generation time point was indistinguishable from an EST1 strain, an est1-Δ rad52-Δ strain displayed an obvious growth defect by 25 generations and was completely inviable by ∼50 generations (depicted schematically in Figure 4). This strongly suggested that recombination was contributing to telomere function even in a newly generated telomerase-defective strain, during the initial period of telomere shortening.
As a reciprocal test of this premise, Rizki and Lundblad (2001) examined the consequences of genetic alterations that increased the frequency of genetic exchanges between telomeric repeat DNA. This work showed that defects in mismatch repair – which relieve a block to recombination between imperfectly matched sequences – enhanced the proliferative capacity of telomerase-defective strains in both S. cerevisiae and K. lactis. Using liquid competition assays, the effects of a mismatch repair mutation were detected during the very early stages of growth of a telomerase-defective yeast strain, even before a senescence phenotype was evident (Figure 4). Increased survival rates were shown to be due to elimination of the anti-recombination activity of the mismatch repair pathway, rather than an indirect effect of increased mutation rates. Notably, enhancement of survival of telomerase-defective K. lactis cells was also dependent on elimination of mismatch repair. Since K. lactis telomeres are composed of perfect telomeric G-rich repeats, this strongly suggested that recombination between partially mismatched sub-telomeric regions could contribute to survival in the absence of telomerase.
More direct support for the potential contribution of recombination between sub-telomeric repeats to telomerase-independent survival came from studies by McEachern and Iyer (2001). Recombination between telomeres was monitored by following phenotypically silent changes in the telomeric repeat tract or the URA3 gene placed in the sub-telomeric region of one telomere. In a telomerase-defective strain, recombination-mediated exchanges were greatly elevated, by up to 1000-fold. Notably, the URA3 gene exhibited rapid dispersal to most, or even all, of the 12 telomeres, reminiscent of the amplification of Y′ elements observed in S. cerevisiae survivors (although note that in these experiments, URA3 was not bracketed by telomeric repeats). Even in cells in which telomerase was still partially active (i.e. telomeres were shortened, but not sufficiently to confer a senescence phenotype), gene conversion rates were increased by 100 to 200-fold, again arguing that recombination at telomeres can be substantially enhanced before chromosome ends become terminally short. The authors proposed that the enhanced recombination rates of stably short telomere mutants reflects a weakened ability of a short telomere to provide a telomeric ‘cap’ that protects chromosome ends from DNA repair activities that normally act on double strand breaks.
However, a strikingly different conclusion was reached when the telomeric changes occurring during the appearance of type II survivors were analysed in detail (Teng et al., 2000). Monitoring of telomere length at multiple time points during the outgrowth of a tlc1− strain in culture showed that telomeres gradually shortened during the pre-survivor period, as previously observed for a telomerase defect, and then abruptly lengthened; this correlated with a sudden switch to an improved growth rate of the culture. When an individually marked telomere in an already established type II survivor was monitored, the previously observed continuous shortening of type II telomeres was interspersed with abrupt elongation events, increasing in size in 1 to 2 kb segments of presumably telomeric repeat tract DNA. Based on these observations, Zakian and colleagues argued against an incremental process of telomere lengthening and instead proposed that a one-step event was responsible. They speculated that the substrate for this event could be extra-chromosomal telomeric repeat DNA, generated by intra-chromatid recombination.
One simple way to reconcile these two sets of observations is to consider a model in which there is a graded increase in recombination that correlates with the severity of the telomeric length defect. In other words, shortened telomeres exhibit enhanced susceptibility to recombination-mediated exchanges, but critically short telomeres are even more recombinogenic. Notably, in the study by McEachern and Iyer (2001), greatly elongated telomeric tracts were not observed in short telomere strains with partial loss of telomerase function, despite the increased telomeric gene conversion rate. This contrasts with the greatly elongated telomeric tracts displayed by survivors of a telomerase null mutation (McEachern and Blackburn, 1996). However, one alternative to a graded recombination response is that partially uncapped telomeres are healed by a different recombination mechanism than that utilized by critically short telomeres.
A summary of potential molecular mechanisms
Work from several different labs in the last several years has expanded our understanding of the genetic requirements for telomerase-independent telomere maintenance in yeast. The first step in this direction came from collaborative work from the Greider and Haber laboratories that investigated the effects of a large set of DNA recombination mutations on the ability of a tlc1− strain to form survivors (Le et al., 1999). Their results demonstrated that at least two different recombination pathways, defined by RAD50 and RAD51, were involved in telomerase-independent survival. Both rad50− tlc1− and rad51− tlc1− strains gave rise to survivors, but a triply mutant rad50− rad51− tlc1− strain was as defective as a rad52− tlc1− strain at generating post-senescent survivors. These two pathways correlated with the two classes of survivors recovered from S. cerevisiae: type I survivors required RAD51 function whereas type II survivors were dependent on RAD50 function (Teng et al., 2000; Chen et al., 2001). However, the correlation between recombination pathway and survivor class may not be absolute, as type II survivors could form in a rad50− mutant background, at least at low frequency, but could not be stably maintained (Chen et al., 2001). Additional analysis has shown that other members of the RAD51 epistasis group (RAD54, RAD55 and RAD57) were also required for the appearance of type I survivors (Le et al., 1999; Chen et al., 2001). Based both on information about genetic requirements and observations of the structural alterations that occur at telomeres, four general models have been proposed for how recombination can maintain telomeres. The next four sections discuss these models.
Break-induced replication (BIR)
BIR is a one-ended recombination event that initiates a replication fork (Figure 1). Strand invasion of an intact chromosome by the broken end (via regions of homology shared between the reporter and donor chromosomes) is followed by replicative copying of donor sequences. If replication continues to the end of the donor chromosome, this will result in the nonreciprocal transfer of chromosomal DNA – as well as terminal telomeric DNA sequences – to the site of the double strand break. Thus, through BIR, a broken chromosome can be healed by capturing a telomere from the donor (Bosco and Haber, 1998). Note that a key feature of this replicative mechanism is that the broken chromosome is not repaired at the expense of the donor chromosome, as would be the case if telomere capture occurred as the result of a nonreciprocal crossover. For a more complete discussion of BIR, see Kraus et al. (2001).
Direct demonstration that BIR can be utilized as a response to a telomerase defect has come from the recent work by Greider and colleagues (Hackett et al., 2001). Their analysis of the molecular events occurring in late generation est1− cultures showed that not only telomeric repeat DNA was lost: much more substantial terminal deletions, extending for many kilobases, also occurred. If left unrepaired, these extensive terminal deletions would presumably lead to lethality, either due to the loss of essential genes or deleterious effects on chromosome transmission. However, these deletions could be healed by break-induced replication, via exchanges with other chromosomes through short stretches of homology shared by the donor and terminally deleted chromosomes. This indicates that a primary mechanism for the generation of telomerase-independent survivors relies on BIR. Although the BIR events recovered by Hackett et al. (2001) involved exchanges between non-homologous chromosomes (because of their experimental design, which dictated what they recovered), healing by BIR can also occur through intrachromosomal translocations (Bosco and Haber, 1998) or unequal sister chromatid exchange.
Multiple studies by the Haber laboratory, using HO-induced double strand breaks, have defined the genetic requirements for BIR, as well as enhancer sites that can facilitate BIR events (Signon et al., 2001; Malkova et al., 2001; Kraus et al., 2001). These studies have demonstrated that a RAD51-independent pathway can promote BIR, using non-random sites to initiate the process. However, RAD51-dependent pathway(s) for BIR are also available (discussed in Kraus et al., 2001). This suggests that survivors that exhibit extensive amplification of sub-telomeric elements (type I) proceed by a RAD51-dependent BIR pathway (Figure 1a), whereas survivors that expand the terminal telomeric tract (type II) are promoted by a RAD51-independent, RAD50-dependent process (Figure 1b). Thus, both classes of survivors may establish and/or maintain their telomeres through BIR.
However, one genetic observation about the requirements for BIR that does not correlate well with the requirements of the survivor pathway is the role of the Sgs1p helicase. Although Sgs1p is not required for conventional BIR (Signon et al., 2001), the appearance of type II survivors is eliminated in telomerase-defective strains that also lack SGS1 (Huang et al., 2001; Cohen and Sinclair, 2001; Johnson et al., 2001). One possible explanation, proposed by Louis and colleagues (Huang, et al. 2001) for the telomere-specific SGS1 requirement may be due to a need to disrupt G-quartets and other secondary structures that single-stranded G-rich telomeric DNA can form.
Integration of extrachromosomal DNA
Integration of extra chromosomal DNA could provide an alternative mechanism for sudden alterations in the size of telomeric and sub-telomeric tracts, via excision and reintegration of chromosomal DNA, as an interchromosomal recombination process (Figure 3b). If the excised DNA also contains an origin of replication, as has been demonstrated for Y′ elements (Chan and Tye, 1983), this provides a rapid means of expansion and dispersal of both sub-telomeric and telomeric DNA. This may provide either an alternative, or even an additional, route to the production or maintenance of type I survivors, in addition to the BIR pathway discussed above. Reintegration by recombination between telomeric DNA present on the Y′ circle and terminal telomeric tracts on the recipient chromosome, as depicted in Figure 1b, would also provide a means of replenishing the chromosome terminus. Notably, in an S. cerevisiae strain that has few Y′ elements, with no tandem arrays or bracketing internal telomeric repeat tracts, type I survivors are rare (Huang et al., 2001), possibly due to the inability to generate extrachromosomal Y′ circles.
Whether excised DNA consisting of simple telomeric DNA is also capable of participating in recombination-mediated reintegration events is unknown. However, Lustig and colleagues (Bucholc et al., 2001) have proposed a model to explain rapid shortening of elongated telomeric repeat tracts, termed TRD (Telomeric Rapid Deletion), such that a potential product of this model is excised telomeric DNA. Since rapid truncations are observed in type II survivors (Teng and Zakian, 1999), this argues that TRD is probably operative in type II survivors strains. If reintegration of telomeric DNA can contribute to survivor formation, this may be a mechanism employed by type II survivors, possibly as a maintenance mechanism of this survivor state once formed. Intriguingly, human ALT cells, which appear to maintain telomeres by a non-telomerase mechanism, contain a novel variant of promyelocytic leukemia (PML) bodies that is composed of telomeric DNA, as well as telomere binding proteins and proteins involved in DNA repair and replication (including Rad51p and Rad52p) (Yeager et al., 1999). However, whether ALT-associated PML bodies are an obligatory intermediate in either the establishment or maintenance steps of the ALT pathway, rather than a byproduct of elongated telomeres, has not yet been resolved.
Rolling circle replication
Rolling circle DNA replication has received increasing attention as a potential mechanism for rapid expansion of the G-rich telomeric tract (Figure 5a). Invasion of a telomeric tract by an extrachromosomal telomeric circle could potentially establish a replication fork that initiates rolling circle replication, thereby allowing extensive elongation of the telomeric tract. Such a model is particularly appealing when considering mechanisms that would promote the abrupt elongation observed during the initial appearance of type II survivors. As pointed out by Teng et al. (2000), conventional gene conversion events between critically short telomeres does not provide an easy explanation for the sudden elongation observed at the termini of type II telomeres, whereas rolling circle replication could provide one potential route for rapid elongation.
Rolling circle replication has been well characterized as a replicative mechanism used by small genomes such as single-stranded bacteriophages and bacterial plasmids (for a review, see Novick, 1998). These replicons depend on specific initiator proteins which both melt open and nick the initiator region, thereby providing a 3′ OH that can serve as a primer for the initiation of DNA synthesis. If rolling circle replication does in fact occur at telomeres, the 3′ terminus of the G-rich strand of the telomere could provide priming activity. However, this model leaves open many questions, such as what activities are required to initiate the process (which involves melting of both the duplex telomeric tract and the invading duplex extrachromosomal DNA, regulation of assembly of the replisome, etc.). In addition, proving that recombination-mediated elongation is templated by a single extrachromosomal telomeric DNA circle (Figure 5a), as opposed to integration of multiple extrachromosomal circles (Figure 3b), will be experimentally challenging.
Elongation of a t-loop
t-loops provide an alternative to the rolling circle replication model (Figure 5b). t-loops are intrachromosomal structures that are formed when the 3′ single-strand G-rich extension at the end of the chromosome invades the duplex telomeric region (Griffith et al., 1999; Munoz-Jordan et al., 2001). These structures have been proposed to provide an architectural means of protecting chromosome ends from degradative and/or DNA repair activities, by masking the telomeric single strand terminus. In addition, such structures potentially regulate elongation by telomerase, since a sequestered 3′ terminus would not be accessible to telomerase. The 3′ terminus of the invading G-rich telomeric strand could also potentially initiate DNA replication, thereby providing a telomerase-independent means for elongation of this terminus. Formation of t-loops does in fact resemble a recombination intermediate, and bears substantial resemblance to the interchromosomal BIR process depicted in Figure 1b. Furthermore, Rad50p (along with other components of the Rad50/Mre11/Nbs1 protein complex) have been invoked as players in either the formation and/or maintenance of t-loop structures (Zhu et al., 2000), suggesting that t-loops might be the initiating structure for type II survivor formation. However, although the structure of a t-loop is certainly suggestive of a replicative model for extension of the chromosome terminus, this presents somewhat of a paradox: t-loops have been proposed as a mechanism to provide chromosome end protection (Griffith et al., 1999), whereas recombination-mediated telomere maintenance is thought to occur in response to a loss of end protection.
Although t-loops have not been identified so far in yeast, detection of these telomeric structures in evolutionary unrelated organisms such as mammals, ciliates and trypanosomes suggests that this is a conserved feature of eukaryotic telomeres. However, although trypanosome t-loops are small (∼1 kb) relative to the average size of mammalian t-loops, this still substantially exceeds the size of telomeric tracts in yeast, particularly the even shorter tracts that occur when telomerase is lacking. Therefore, a potential role for t-loops in rescuing yeast senescence and promoting survivors remains speculative at this time.
Potential parallels to an ALT (Alternative Telomere Lengthening) pathway in human cells
As mentioned at the start of this review, interest in the yeast survivor pathway has been enhanced by the possibility that recombination may contribute to telomere maintenance in human cells as well. If so, this may have particular significance to processes that promote neoplasia. In fact, although most immortalized human cells, including tumor cells, have substantially up-regulated telomerase (Shay and Bacchetti, 1997), a subset of immortalized cells do not display high levels of telomerase activity (Bryan et al., 1995, 1997). Furthermore, telomeres in these cells show a strikingly different profile, displaying heterogeneous telomeres that range from very short to lengths that are elongated by up to 50 kb. The absence of detectable telomerase enzyme activity, coupled with the striking telomeric profile, has led to considerable speculation that ALT may proceed by the same mechanism(s) that promote telomere elongation in K. lactis survivors or type II survivors in S. cerevisiae.
The specific genetic requirements that promote the ALT pathway have not yet been elucidated, due to the difficulties of genetic analysis in human cells, nor has an inducible system for ALT, where the intermediate steps in the pathway can be analysed, been established. In addition, the steady attrition of telomeric DNA observed in K. lactis and type II S. cerevisiae survivors have not been observed. However, numerous studies have shown that human chromosome termini are subject to enhanced levels of recombination, as monitored by sequence polymorphisms in sub-telomeric regions (see Wilkie et al., 1991; Baird et al., 2000 for several examples). Additional work indicates that terminal deletions found on human chromosomes can be healed by capture of sub-telomeric and telomeric sequences from another chromosome (Meltzer et al., 1993). A study by Reddel's laboratory (Dunham et al., 2000) was designed to ask specifically whether ALT telomeres participated in recombination, by monitoring the transmission of a selectable marker inserted into telomeric sequences in either ALT or telomerase-plus immortalized cells. This marker spread to other chromosome ends much more frequently in ALT cells than in telomerase-plus cells, suggesting that the telomeres in ALT cells (or a subset, such as the shortest telomeres?) were more recombinogenic than telomeres maintained by telomerase. Notably, when the same marker was placed in a sub-telomeric location, dispersal to other chromosomes was not observed, in contrast to the results obtained in K. lactis by McEachern and Iyer (2001).
It is also worth noting that recombination may contribute to telomere maintenance and potentially tumor formation, even without conversion to the full ALT pathway. In particular, any genetic alteration that enhances recombination rates at telomeres – such as defects in mismatch repair – may partially relieve the telomere maintenance barrier to proliferation, thereby contributing to the mechanism of oncogenesis.
Can the ALT pathway and telomerase-mediated telomere maintenance co-exist in the same cell?
A number of recent papers have been directed at the above question, by examining the consequences of reconstitution of telomerase activity in ALT cells as a result of ectopic expression of telomerase subunits (Perrem et al., 2001; Cerone et al., 2001; Ford et al., 2001; Grobelny et al., 2001) or by fusion of ALT cells with telomerase-positive tumor cell lines (Perrem, et al. 2001). However, the conclusions from these four groups were somewhat contradictory. Two groups reported that when telomerase was ectopically expressed, short telomeres were preferentially elongated, but the ALT pathway was not repressed (Grobelny et al., 2001; Cerone et al., 2001). In contrast, Shay and Wright and colleagues, using the same ectopic expression approach, observed rapid reduction of the longest telomeres and concluded that ALT was inhibited (Ford et al., 2001). Reddel and co-workers did not observe inhibition of ALT when telomerase was ectopically expressed, but when ALT cells were fused with telomerase-positive tumor cells, telomeres showed rapid initial decreases in length, followed by more gradual shortening (Perrem et al., 2001).
Observations from yeast may provide some insight into the potential complexities of these experiments. The genomic alterations that occur in telomerase-defective survivors can be very stable, even when telomerase is restored. Introduction of EST1 back into an est1-Δ survivor that exhibits extensive Y′ amplification results in elongation of the terminal telomeric tract back to wild type length, as well as a substantial reduction in Y′ copy number (Lundblad and Blackburn, 1993; Teng and Zakian, 1999). However, even with extensive propagation, the distribution of Y′ elements does not return to its starting point: each telomere retains at least one Y′ element in an EST1 strain that has a previous history as a telomerase-defective survivor (V Lundblad, unpublished results). When this strain is converted back to est1-Δ, it again exhibits telomere shortening and senescence, but in this newly senescent strain, survivors appear more readily and at earlier time points. This suggests that the retained single Y′ elements prime the re-appearance of survivors. Similarly, following reintroduction of telomerase into a type II survivor, telomeres return to a roughly wild type telomeric restriction pattern only very slowly (Teng and Zakian, 1999). This suggest that even if telomeres are no longer highly recombinogenic after the reintroduction of telomerase, the stable extended telomeric tracts still decay only very slowly.
Extending these observations to human cells, this suggests that using telomere length as a monitor of whether a recombination-promoted pathway for telomere maintenance is still operating in the presence of telomerase may not be ideal. This monitors the product, rather than the presence of an active mechanism. The solution may be to develop assays that directly monitor the frequency of recombination events at telomeres. One such assay, using a reporter gene that marks a telomeric tract, has been described above (Dunham et al., 2000). An alternative approach relies on the presence of natural variants in the telomeric repeat tract. Human telomerase synthesizes TTAGGG repeats onto chromosome termini (Morin, 1989). However, the proximal end of this repeat tract contains variant telomeric repeats such as TGAGGG and TCAGGG, which can be monitored by a PCR-based assay (Baird et al., 1995). Since the distribution of these variant repeats is highly variable, this assay may provide a semi-quantitative means of monitoring the extent of recombination at human chromosome ends.
Summary and future possibilities
Experiments using budding yeasts have provided both specific models and genetic requirements that dictate the frequency of recombination-dependent exchanges between telomeres in cells that lack telomerase. These observations have implicated specific recombination proteins and also highlighted other alterations, such as defects in mismatch repair, that may similarly influence the survival of human cells that fail to express telomerase. Additional studies, not discussed in this review, suggest that alterations in components of telomeric chromatin can also influence either the types of recombination events that occur at the telomere (Teng et al., 2000) or the frequency with which the survival pathway appears (Chandra and Lundblad, unpublished observations). All of these examples in yeast provide a possible roadmap for experiments that may help elucidate the ALT pathway.
I thank Jim Haber and Jenny Hackett for explaining the subtleties of the genetic requirements of BIR to me, Sara Evans for excellent editorial assistance and figure preparation, and Lou Zumstein, Rachel Cervantes and Erin Pennock for critical reading of the manuscript. Work in the author’s laboratory is supported by grants from the NIH and the Ellison Medical Foundation.