We determined the mechanisms that lead to chromosome stabilization after telomere loss in Saccharomyces
. Three pathways were able to stabilize broken chromosomes: homologous recombination, de novo telomere addition, and nonhomologous end joining (Figure ). Regardless of the mechanism, each stabilization event resulted in acquisition of a telomere. Although previous work from our lab demonstrated that loss of a telomere often results in loss of the affected chromosome (Sandell and Zakian, 1993
), it was not clear from this or other previous studies (Kramer and Haber, 1993
) whether stabilization ever occurred without telomere acquisition. Thus, even although double-strand breaks in yeast bind several telomeric proteins, such as the Ku heterodimer and Sir proteins (Martin et al., 1999
; Mills et al., 1999
), unlike the case for HP1 binding in Drosophila
, this binding is not sufficient to confer telomere function in the absence of telomeric DNA. Together with previous studies (Kramer and Haber, 1993
; Sandell and Zakian, 1993
), these data argue that telomeric DNA is required for the stable maintenance of linear chromosomes in Saccharomyces
. This dependence is likely to apply to other eukaryotes, which, like yeast, maintain their telomeres by a telomerase mechanism. Because telomeric DNA in Drosphila
is not telomerase generated (Biessmann and Mason, 1997
's ability to maintain telomere function without a special telomeric DNA may be related to its unusual mechanism of end replication.
As expected, in a wild-type cell, the majority of the stabilized test chromosomes acquired a new telomere by homologous recombination with the endogenous copy of chromosome VII (Supplementary Table 1). However, even in the rad52
strain, most (80%) stabilization events involved homologous recombination. Surprisingly, these events were not eliminated in a strain with a complete deletion of RAD52
(rather than the insertion allele used for most experiments) nor in a rad52
strain that also lacked Rad59p, a Rad52p-related protein whose function partially overlaps that of Rad52p (Bai and Symington, 1996
; Bai et al., 1999
) (Teng and Zakian, unpublished data). Extremely rarely, occurring at <0.003% of the broken chromosomes and accounting for <1% of the stabilization events, stabilization was due to nonhomologous end joining (Figure E). In contrast, in mammalian cells, at least half of induced double-strand breaks are repaired by nonhomologous recombination (Liang et al., 1998
Although 20% of the stabilization events in the rad52
strain were due to telomere addition (Supplementary Table 1A), these events were relatively rare, stabilizing <0.1% of the broken chromosomes. Indeed, healing a broken chromosome in yeast by de novo telomere addition is so rare that a previous study that analyzed fewer cells did not observe any such events, leading the authors to speculate that de novo telomere addition does not occur on natural yeast chromosomes (Kramer and Haber, 1993
). Telomere addition anywhere in the ~220-kb region between the telomere and LYS5
would have been detected in our assay. Nonetheless, telomere formation occurred at a very limited number of sites. Thirty percent of these events were within a 300-base pair portion of URA3
, <1 kb from the site of chromosome breakage (Supplementary Table 1B and Figure , A and B). URA3
had one 9- and one 11-base pair tract of telomere-like DNA, separated by only 230 base pairs (Figure B, bold and underlined). Telomere-like tracts of this length spaced this close together were relatively rare in the yeast genome. Most (70%) telomere addition events occurred ~50 kb from the end of the test chromosome (Supplementary Table 1B). Six of six sequenced ~50-kb terminal deletions were within a few kilobases of a (CA)17
tract (Figure A, closed symbols). Four other telomere addition events were mapped to this region by Southern hybridization (Figure A, open symbols). After the (CA)17
tract, the next longest tract of telomere-like DNA in the ~220-kb region between the telomere and LYS5
on chromosome VII was a 14-base pair tract at position 76, 214 base pairs (Figure B). Based on the absence of telomere additions events near this 14-base pair tract (Figure A), we estimate that telomere addition was ≥10 times more frequent near the (CA)17
tract than near the 14-base pair tract. However, other features of a telomere-like tract, in addition to its length, such as its chromosomal context, might affect its ability to promote telomere addition.
Tracts of telomere-like DNA as long as the (CA)17
tract were extremely rare in nontelomeric regions of the yeast genome. For example, seven yeast chromosomes had no internal tract of telomere-like DNA longer than 14 base pairs in the proper orientation to promote telomere addition (Figure C). We propose that long tracts of telomere-like DNA, such as the (CA)17
tract on chromosome VII, increase the probability of telomerase-mediated telomere addition. This proposal was also made in a previous study in which a 78-base pair tract of C4
DNA, a ciliate telomeric sequence not found normally in yeast, promotes telomere addition at an HO-induced double-strand break on chromosome III (Kramer and Haber, 1993
). In this system, no telomere addition events are detected in the absence of the ciliate tract, consistent with our finding that chromosome III had no long tract of telomere-like DNA (Figure C). This model suggests that de novo telomere addition is rare in wild-type cells in part because there are few internal tracts of telomere-like DNA long enough to promote telomere addition.
How might long tracts of telomere-like DNA increase the probability of telomerase elongation? After HO induces a double strand break, the 5′ end is degraded much faster than the 3′ end (White and Haber, 1990
). This degradation will expose the G-rich strand of telomere-like tracts (Figure ). Because there was a minimum of 37 bps between the (CA)17
tract and the start of the newly added telomere (Figure , C and D), the (CA)17
tract was not itself the site of telomere addition. Previous studies also found that several base pairs can intervene between the site of telomere addition and a telomere-promoting sequence (Murray et al., 1988
; Wang and Zakian, 1990
; Kramer and Haber, 1993
), but our analysis is the first to document the ability of a naturally occurring tract of chromosomal, telomere-like DNA to promote telomere addition.
Figure 6 Model for de novo telomere addition after chromosome breakage. The striped and spotted rectangles represent, respectively, the C and G-strands of a long stretch of telomere-like DNA, such as the (CA)17 tract on chromosome VII. Small closed circles represent (more ...)
These data argue that telomere-like tracts do not act simply by increasing the probability of primer pairing to the template region of telomerase RNA. Rather, we suggest that telomere-like tracts recruit telomerase by interacting with the anchor site (or another site) on telomerase (Figure ). To explain the processivity of ciliate and human telomerase in vitro, DNA primers are proposed to make contact with telomerase at two sites: the catalytic site where the primer is base paired to telomerase RNA, and the anchor site (Greider, 1991
; Morin, 1991
; Hammond et al., 1997
). Binding at the anchor site can retain telomerase at the telomere during the translocation step, when the primer is released from the active site. Consistent with a model in which the single strand (TG)17
tract binds telomerase, TG1–3
but not C1–3
A or duplex oligonucleotides bind Saccharomyces
telomerase in vitro (Lue and Xia, 1998
). Yeast telomeres are clustered in vivo (Gotta et al., 1996
). This clustering might facilitate the ability of a shorter tract of telomere-like DNA that is close to a telomere, such as the 11-base pair tract within URA3
(Figure B), to recruit telomerase from another telomere, whereas the same length tract far from a telomere would be less effective.
Another reason telomere addition at double-strand breaks is rare is that it is inhibited by the Pif1p DNA helicase. When cells lacked the Pif1p helicase, the frequency of telomere addition increased almost 200-fold (Supplementary Table 1A). This increase probably underestimates the inhibitory effects of Pif1p because a pif1-m2
strain does not have as severe a telomere phenotype as a pif1Δ
strain (Schulz and Zakian, 1994
). In addition, the distribution of telomere addition sites was altered in the absence of nuclear Pif1p (Supplementary Table 1B). First, a larger fraction (78%) of the telomere addition events were within 1 kb of the telomere in the rad52 pif1-m2
strain than in the rad52
strain (30%) (Supplementary Table 1B). The most common site for telomere addition in the rad52 pif1-m2
strain, used in 30% of the stabilization events, was only 9 bps from the HO cut site (Figure C, site 1). At this site, new telomeres were added directly to a CCCA tract, the very first telomere-like DNA internal to the HO cut site (Figure D, site 1). Telomere addition at this site was never detected in the presence of Pif1p, even although we searched specifically for these events (Figure E).
The distribution of de novo telomere addition events internal to URA3 was also affected by Pif1p (compare Figure B, rad52 pif1 to A, rad52 strain). Whereas all nine of the large terminal deletions in the rad52 strain were within a few kilobases of the (CA)17 tract, only one of eight such events was near this tract in the rad52 pif1 strain. Nonetheless, telomere addition was not random in pif1 cells because in 23 of the 24 sequenced telomere addition events, the new telomere was added to a 3–9-base pair stretch of telomere-like DNA (Figure , B and D). A similar spectrum of addition site sequences was seen in the rad52 strain where 7 of 10 new telomeres were added at 3–11-base pair tracts of telomere-like DNA (three were added to nontelomeric DNA; Figures B and , C and D).
How does Pif1p inhibit the frequency of telomere addition and alter the distribution of these events? Because its catalytic activity is necessary for Pif1p to inhibit lengthening of existing telomeres (Zhou et al., 2000b
), the simplest possibility is that Pif1p also acts as a helicase when it affects telomere formation at a double-strand break. We suggest that telomerase interacts with DNA breaks by base pairing of telomerase RNA to very short stretches of telomere-like DNA, stretches that are very common in yeast DNA (Figure B). However, in PIF1
cells, these interactions are rarely productive because Pif1p dissociates them (Figure , left). Thus, in wild-type cells, telomerase action at a double-strand break requires an additional mechanism, such as a long stretch of telomere-like DNA that can retain telomerase after Pif1p-mediated displacement of its active site from the DNA primer (Figure , right). This model explains both the increased frequency and the altered distribution of addition events in pif1
cells because in the absence of Pif1p, telomerase would have a higher probability of productive association with very short tracts of telomere-like DNA.
Some chromosomal rearrangements isolated from both normal human fibroblasts and human tumor cells, once thought to be terminal deletions, have been found to be subtelomeric translocations (Meltzer et al., 1993
). These data suggest that de novo telomere addition is much rarer in human cells than originally thought. If the human Pif1p-like protein (Zhou et al., 2000b
), like its Saccharomyces
counterpart, inhibits de novo telomere addition, it could contribute to genome stability by limiting the number of terminal deletions, thereby reducing events that result in loss of heterozygosity.