The results presented here demonstrate for the first time that chromosome healing in mammalian cells can occur through more than one mechanism. One mechanism involves the de novo
addition of telomeric repeat sequences by telomerase, as demonstrated by the fact that while chromosome healing is a common event in wild type ES cells (22 of 35 events), chromosome healing rarely occurs in the 10PTKO-H and 10PTKO-4B ES cell lines that are deficient in telomerase (1 of 62 events). The one chromosome-healing event that was observed in these two TERT
knockout cell lines involved the insertion of a 23 bp sequence between the site of the DSB and the telomeric repeat sequences. The insertion of a small DNA fragment suggests a mechanism involving recombination, since small insertions are common at junctions repaired by NHEJ [26
In addition to the mechanisms mentioned above, our results also show that some TERT
knockout ES cell lines can efficiently perform chromosome healing through a telomerasein-dependent mechanism. This mechanism is active in the 10PTKO-A ES cell line, which unlike the TERT
knockout ES cell lines 10PTKO-4B and 10PTKO-H, is very adept at chromosome healing despite the absence of telomerase activity (20 of 34 events). However, this mechanism does not appear to be active in ES cells with wild-type telomerase. In contrast to the wild-type ES cell lines, chromosome healing in 10PTKO-A was never observed to initiate from the terminal nucleotide of the 4 bp overhang generated by I-Sce
I (site a, ), which was common in wild type ES cell lines. Instead, telomeres were often added at the site with 2 bp of homology (site c), which was rarely used in wild type ES cell lines. Therefore, although this alternative mechanism also utilizes microhomology for chromosome healing, its requirements for telomere addition are different from the telomerase-dependent mechanism. In view of the absence of telomerase activity, it is likely that chromosome healing in 10PTKO-A involves the addition of preexisting telomeric repeat sequences to the site of the DSB. This could occur by NHEJ, which commonly utilizes microhomology for end rejoining and the deletion of a small number of nucleotides [51
], both of which were observed at sites of chromosome healing in 10PTKO-A. Consistent with our results with 10PTKO-A, we previously demonstrated the addition of telomeric repeat sequences to both ends of a plasmid sequence transfected into an immortal human cell line [32
] that was later shown to maintain telomeres through a telomerasein-dependent mechanism [33
]. Similar to the present study, this addition of telomeric repeat sequences also involved microhomology, with 2 bp of microhomology at one end of the plasmid, and 3 bp of microhomology at the other end.
An important difference between the chromosome healing in the wild type and telomerase-deficient ES cell lines is the length of newly added telomeres. The newly added telomeres in the wild type ES cell lines in this study are already 75 kb at the earliest time of analysis, similar to the length of the telomere in the parental ES cell lines. We estimate that less than 50 cell divisions had occurred during the period in culture required to grow sufficient cells to isolate genomic DNA for Southern blot analysis. Thus, an average of approximately 1,500 bp would have been added to the new telomere with each cell division, although initially the rate of telomere elongation may be even greater, since we previously observed that at later times the newly added telomeres can grow much more gradually [31
]. Regardless, these results demonstrate that although mouse telomerase has poor processivity in cell lysates [53
], it can be highly processive during the restoration of a lost telomere in living cells.
In contrast to ES cell lines with wild type telomerase, the newly added telomeres in the telomerase-deficient 10PTKO-A ES cell line are relatively short, with some being as short as 1 kb in length at the time of analysis. In addition, rather than being elongated, the telomeres shorten during passage in culture, which eventually results in the loss of the subtelomeric plasmid sequences in many cells in the population. However, the telomere is elongated in other cells, demonstrating that the 10PTKO-A ES cell line is capable of maintaining telomeres through a telomerase-independent pathway, although this pathway is relatively inefficient in protecting critically short telomeres compare to human cells that maintain telomeres through a telomerase-independent pathway [33
]. This mechanism is not normally found in ES cell lines, because other studies have demonstrated that telomerase-deficient ES cell lines lack the ability to preferentially elongate shortened telomeres [54
]. The acquisition of an alternative mechanism for telomere maintenance has previously been described for telomerase-deficient mouse ES cells [47
]. The analysis of newly added telomeric DNA in one of the telomerase-deficient ES cell lines in this earlier study demonstrated that it was composed of both telomeric and subtelomeric DNA [31
]. Although the newly added telomeres found at the site of chromosome healing in our study invariably consisted of telomeric repeat sequences, it cannot be ruled out that the telomeres added by chromosome healing or subsequent elongation also contain nontelomeric DNA sequences. This previous study also found that ES cell clones utilizing telomerase-independent telomere maintenance only appeared following a telomere-shortening-induced crisis after 450 cell divisions in culture. Why the 10PTKO-A ES cell line would have acquired an alternative mechanism for telomere maintenance after only a relatively few generations in culture is not known. One likely possibility is that it initially had short telomeres due to telomere shortening that occurred in vivo
, which would have selected for cells that were capable of maintaining telomeres by alternative mechanisms. This telomere shortening in vivo
could have occurred because the ES cell lines used in our study were derived from heterozygous TERT
knockout mice. Mice that are heterozygous for TERT
knockout experience gradual telomere shortening with each generation [55
], although unlike homozgous TERT
knockout mice, the heterozygous mice are able to maintain short telomeres [56
]. Consistent with telomere shortening in vivo
, two of the three TERT
knockout cell lines used in our study had telomeres that were much shorter than telomeres in ES cells with wild type TERT.
Another important feature of chromosome healing is that in all but one subclone with a 25 bp deletion, it occurs at the location of the I-SceI-induced DSB. This contrasts with sister chromatid fusion, in which degradation of one or both sister chromatids commonly occurs. This degradation prior to sister chromatid fusion can be extensive, in one subclone reaching 30 kb in length. Why chromosome healing only occurs at the site of the initial I-SceI-induced DSB, and not at proximal sites following DNA degradation, is not known. However, it is not due to selection for the proximal puro gene, since the chromosome healing observed with selection with ganciclovir alone also occurred at the I-SceI site. Chromosome healing may therefore be in competition with other DNA repair mechanisms, which are associated with processing and degradation of DNA at the break site. If these other DNA repair enzymes encounter the DSB first, then telomerase or the telomerase-independent mechanism would be prevented from accessing the free end for telomere addition.
The fact that chromosome healing is a common event in our assay system would appear to contradict other studies that have concluded that chromosome healing is not a common event at DSBs generated by either I-Sce
I or ionizing radiation [26
]. Even when chromosome healing is observed, the evidence suggests that it usually involves the capture of the ends of other chromosomes [57
]. One possible explanation for the prevalence of chromosome healing at the I-Sce
I-induced DSB in our system is the proximity of the telomere. Studies in yeast have demonstrated that de novo
telomere addition is much more likely to occur near pre-existing telomeric repeat sequences, which is independent of which side of the break the telomeric repeat sequences are located [19
]. De novo
telomere addition at DSBs at interstitial sites may also be inhibited as it is in yeast, where its inhibition by Pif1 is proposed to be a mechanism for preventing chromosome healing from interfering with DSB repair [21
The different outcome of DSBs near telomeres compared to DSBs at most other locations may also reflect the fact that the proximity to a telomere may influence DSB repair. DSBs near telomeres in yeast have been show to be poorly repaired by NHEJ, and therefore result in complex chromosome rearrangements [34
]. This deficiency in repair in subtelomeric regions may be due to the role of the telomere in preventing chromosome fusion, since telomeric repeat sequences in yeast have been shown to suppress the activation of cell cycle checkpoints in response to DSBs [35
]. Similarly, the human TRF2 protein, which is required to prevent chromosome fusion, has been demonstrated to inhibit ATM, whose activation is instrumental in the cellular response to DSBs [36
]. Alternatively, a deficiency in the repair of DSBs within subtelomeric DNA could also result from the heterochromatic structure of these regions ([58
], unpublished results). Regardless of the cause, a deficiency in NHEJ within subtelomeric regions would make chromosome healing an important option for the cellular response to DSBs occurring near telomeres.
Our results also show that chromosome healing can prevent chromosome instability resulting from DSBs near telomeres. Because chromosome healing almost always occurs at the site of the break, it must precede and prevent degradation, sister chromatid fusion, and subsequent B/F/B cycles. The importance of sister chromatid fusion and B/F/B cycles resulting from telomere loss in DNA rearrangement in cancer is becoming increasingly evident. Telomerase deficient mice that also have a knockout in p53, and therefore are unable to eliminate cells with genomic instability, have a high frequency of carcinomas that contain chromosomes with rearrangements typical of B/F/B cycles [59
]. In addition, telomere loss has been shown to play an important role in human cancer, and can continue to occur despite the expression of telomerase [62
]. Chromosome healing could therefore limit the extent of the genomic instability commonly associated with human cancer, since chromosome healing has also been observed in human tumor cells in response to spontaneous [14
] or I-Sce
I-induced (unpublished observation) telomere loss. However, like mouse ES cells, the efficiency of chromosome healing is not sufficient to prevent sister chromatid fusion and instability. Knowledge regarding the regulation of chromosome healing may therefore provide future approaches to limiting genomic instability in human cancer cell for the purpose of inhibiting tumor cell progression or adaptation to cancer therapy.