Dynamics of DNA Double-Strand Breaks Revealed by Clustering of Damaged Chromosome Domains

See allHide authors and affiliations

Science  02 Jan 2004:
Vol. 303, Issue 5654, pp. 92-95
DOI: 10.1126/science.1088845


Interactions between ends from different DNA double-strand breaks (DSBs) can produce tumorigenic chromosome translocations. Two theories for the juxta-position of DSBs in translocations, the static “contact-first” and the dynamic “breakage-first” theory, differ fundamentally in their requirement for DSB mobility. To determine whether or not DSB-containing chromosome domains are mobile and can interact, we introduced linear tracks of DSBs in nuclei. We observed changes in track morphology within minutes after DSB induction, indicating movement of the domains. In a subpopulation of cells, the domains clustered. Juxtaposition of different DSB-containing chromosome domains through clustering, which was most extensive in G1 phase cells, suggests an adhesion process in which we implicate the Mre11 complex. Our results support the breakage-first theory to explain the origin of chromosomal translocations.

DSBs induced by DNA damaging agents or during genome duplication are hazardous because interactions between DNA ends from different DSBs can produce tumorigenic chromosome translocations (1). Although much is known about DSB repair, less is known about how ends from different DSBs meet (2). Two theories have been advanced to describe this event. The “contact-first” theory proposes that interactions between unrelated chromosome breaks can take place only when the breaks are created in chromatin fibers that colocalized at the time of DNA damage induction (3). The “breakage-first” theory assumes that breaks formed at distant locations can subsequently be brought together to produce translocations (4). The breakage-first theory predicts that DSBs should move over large distances in the nucleus before interacting. Whether such extensive migration and subsequent interaction of DSBs can actually occur is unclear (57).

By exposing cells to α particles from a radiation source positioned alongside the cells (8), we introduced near-horizontal linear DSB tracks (912). Potential movements of the DSBs relative to each other should result in distortion of their initial pattern. DSBs were visualized by immunodetection of histone H2AX phosphorylation (Fig. 1), resulting in γH2AX, which rapidly expands through chromatin on both sides of a DSB (13, 14). Changes in the spatial distributions of DSBs were studied by analyzing track morphology at various intervals after exposure to α radiation. The linear tracks in HeLa nuclei at 2.5 min after impact closely approximated the trajectories of the α particles through the nucleus. In these tracks, individual chromosome domains containing γH2AX (γH2AX-CD) could be observed (Fig. 1A). The average track length was 10.1 ± 2.8 μm, and the average number of γH2AX foci was 22 ± 2 per 10 μm. Changes in track morphology could be observed within 5 min after exposure when the cells were maintained at 37°C. At 21°C, track modifications were inhibited, indicating that track reorganization is not solely due to passive diffusion. We classified tracks into two main groups on the basis of differences in their morphological changes over time. In one group, γH2AX-CDs along the tracks clustered (Fig. 1, B to E), whereas the other group of tracks contained only nonclustering γH2AX-CDs (Fig. 1, F and G).

Fig. 1.

Changes in spatial distributions of DSBs in time. (A) Distributions of γH2AX-CDs in HeLa nuclei 2.5 min after exposure to α radiation. (B to D) γH2AX-CD distributions in clustering tracks 7.5 min, 15 min, and 60 min after irradiation. (E) Multiple individual γH2AX-CDs, three of which are indicated by an arrow, can be observed in single clusters. (F and G) Distributions of γH2AX-CDs in nonclustering tracks 30 min and 180 min after irradiation. DNA was stained with Hoechst 33342 (blue). DSB-containing chromosome domains were visualized with γH2AX antibodies (red). The small γH2AX foci in the background are indicative of S-phase cells (24). Tracks with similar general characteristics were also observed in normal human primary fibroblasts (table S1) and Chinese hamster ovary cells. Scale bars, 2 μm.

The fraction of tracks with γH2AX-CD clusters is presented in Table 1 as Group I. At 5 to 10 min after exposure, stretches devoid of γH2AX-CDs appeared in these tracks (Fig. 1B), and at 15 min most of the γH2AX-CDs had congregated into small clusters (Fig. 1C). The clustering process continued for more than 60 min (Fig. 1, D and E). Group I tracks could be divided into two subgroups. In tracks of Subgroup Ia, all the γH2AX-CDs were contained in three to five large clusters. No separate γH2AX-CDs were present in these tracks. Subgroup Ib tracks contained a mixture of clustered and separate nonclustered γH2AX-CDs.

Table 1.

Quantification of α particle–induced γH2AX-CD track morphology and Rad51 colocalization in HeLa cells as a function of time after irradiation. Displayed are the percentages of tracks observed within the indicated categories; the absence (–) or presence (+) of Rad51 foci is denoted for each. Tracks in the category “Other” could not be classified unambiguously as Group I or II. At least 100 tracks per time point were analyzed.

Group I: tracks with γH2AX-CD clusters Group II: tracks without γH2AX-CD clusters Other (%)
Time after irradiation (min) Subgroup Ia: clusters only Subgroup Ib: clusters and nonclusters
- + - + Total (%) - + Total (%)
0 0 0 0 0 0 97 3 100 0
15 6 0 0 0 6 73 16 89 5
30 19 0 0 7 26 39 25 64 10
60 38 0 0 12 50 2 31 33 17
120 24 0 0 23 47 0 36 36 17
180 26 0 0 46 72 0 9 9 19

We propose that the formation of γH2AX-CD clusters and the emergence of vacant stretches in Group I tracks are due to movement of γH2AX-CDs. Individual γH2AX-CDs could still be distinguished in the clusters of some Group I tracks (Fig. 1E). The total number of γH2AX-CDs in these clustered tracks was 20 ± 7 foci per 10 μm at 30 min. This is close to the number found in the linear tracks analyzed immediately after irradiation, supporting the argument that clusters are formed by relocation of γH2AX-CDs. Analysis of the tracks revealed that seven or more γH2AX-CDs, separated by distances of several μm, could be brought together to form a cluster.

In contrast to Group I tracks, Group II tracks consisted exclusively of nonclustering γH2AX-CDs (Fig. 1, F and G). The patterns in this group varied from linear to scattered. The scattering of γH2AX-CDs in these tracks demonstrates that γH2AX-CDs can move within a limited range, notwithstanding their attachments to neighboring chromosome domains. We observed nonclustering tracks with γH2AX-CDs that had moved up to ∼2 μm away from the linear track (Fig. 1G). The extent of movement is of the same order as that observed for mobile chromosomal loci that do not contain DSBs (15). The movement of the γH2AX-CDs was sufficient to produce multiple contacts between the γH2AX-CDs in the majority of the tracks, but this rarely led to aggregation of the domains.

Next, we investigated the relation between γH2AX-CD clustering and cell cycle phase by analyzing DSB tracks together with the signal of the DNA replication marker proliferating cell nuclear antigen. The majority of the clustered Group I tracks (83%) was found in G1 phase cells. Most of the nonclustered Group II tracks (67%) occurred in S and G2 phase cells. Given the interval of 60 min between DSB induction and cell fixation, clustering is most prevalent in pre–S-phase cells. The absence of clustering in tracks with mobile γH2AX-CDs (Fig. 1, F and G) indicates that chromatin movement is not the only requirement for clustering. Formation of stable γH2AX-CDs clusters suggests the existence of an adhesive process that binds together multiple γH2AX-CDs.

The adhesion between γH2AX-CDs could involve proteins that stabilize damaged chromosomes by interconnecting chromatin fibers decorated with γH2AX. Potential candidates are proteins of the structural maintenance of chromosome (SMC) family (16). For example, the SMC protein–containing Mre11 DSB repair complex can specifically tether linear DNA molecules (17). A subfraction of the Mre11 complex has properties expected of a protein complex involved in γH2AX-CD adhesion. Of the two types of DNA damage–induced accumulations of the Mre11 complex, one is disrupted by extraction of nuclei with detergent, whereas the other is not (18). We found that the detergent-sensitive fraction was present in a dispersed manner throughout the clustered γH2AX-CDs (Fig. 2A), consistent with a possible role of the Mre11 complex in γH2AX-CD adhesion. By contrast, the detergent-resistant fraction showed only a single small Mre11 focus per γH2AX-CD or a few Mre11 foci in clusters of γH2AX-CDs (Fig. 2, B and C). The oligomers of the Mre11 complex that can form at DNA ends (17) could represent the Mre11 complex signal observed in these foci, whereas the global association of the complex with the γH2AX-CDs could be mediated through the forkhead-associated phosphoepitope binding domain of its NBS1 component (19).

Fig. 2.

Mre11 and γH2AX-CD clustering. (A to C) Distribution of the Mre11 complex in γH2AX-CDs 30 min after exposure to α radiation. Differential extraction of HeLa cells containing α particle–induced γH2AX-CDs (red) reveals two modes of Mre11 complex association with γH2AX-CDs. Dispersedly colocalizing Mre11 (green) is present in cells that have not been treated with detergent. Colocalization of Mre11 and γH2AX results in yellow (A). Avidly bound Mre11 complex, revealed by detergent extraction of HeLa cells, is present in bright foci within γH2AX-CDs [(B) and (C)]. (D) Linear arrangement of Mre11 foci in HeLa cells in a nonclustering track 30 min after α irradiation. (E and F) Morphology of clustering γH2AX-CDs in normal human primary fibroblasts (E) and ATLD primary fibroblasts (F) 15 min after irradiation. Scale bars, 2 μm.

To obtain evidence for involvement of the Mre11 complex in γH2AX-CD clustering, we analyzed track morphologies in primary fibroblasts from ataxia telangiectasia-like disorder (ATLD) patients; these fibroblasts have a reduced level and function of the Mre11 complex (20). Three observations suggest that clustering of α particle–induced γH2AX-CDs is affected in ATLD cells. First, the percentage of tracks with Subgroup Ia morphologies, containing γH2AX-CDs in large clusters only and lacking Rad51 foci, was reduced from 10% in normal fibroblasts to 2% in ATLD fibroblasts at 60 min after irradiation. Second, consistent with less efficient γH2AX-CD clustering in ATLD cells, this reduction was concomitant with an increase from 0% to 15% in tracks with Subgroup Ib morphology, containing both clustered and nonclustered γH2AX-CDs and lacking Rad51 foci. Third, in the ATLD cells that did show clustering of γH2AX-CDs, the clusters were often more extended and branched in appearance compared with their more compact appearance in control cells (Fig. 2, E and F).

For additional information on DSB track clustering, we probed the γH2AX-CD tracks for Rad51, a protein involved in DSB repair by homologous recombination (1). The linearity of the tracks of Rad51 (Fig. 3A) and Mre11 foci (Fig. 2D) observed upon exposure to α particles demonstrates colocalization of the DNA repair proteins with DSBs. Rad51 (Fig. 3, B to E) and avidly bound Mre11 (Fig. 2, B and C) were concentrated mainly in small foci located at the interior or at the periphery of the γH2AX-CD. Moreover, we very rarely observed more than one repair protein focus in an individual γH2AX-CD. Mre11 was present in Subgroups Ia and Ib and Group II tracks in ∼80% of the cells, which suggests that Mre11-mediated DSB processing occurred in all track types. By contrast, Rad51 localized to Subgroup Ib and Group II tracks but was absent from tracks that displayed clustering only (Subgroup Ia; Table 1). Possibly, DSB processing in these clusters proceeds through a Rad51-independent pathway. This is consistent with our observation that the cells containing this track morphology are in the G1 phase of the cell cycle in which Rad51 foci do not form (21).

Fig. 3.

Spatial relations between γH2AX-CDs and Rad51 foci. (A) Linear arrangement of Rad51 foci (green) in HeLa cells in a nonclustering track 30 min after α irradiation. (B to E) Rad51 foci in γH2AX-CDs 30 min [(B) and (C)] and 60 min [(D) and (E)] after irradiation. DNA, γH2AX, and Rad51 are shown in blue, red, and green, respectively. (C) and (E) are three-dimensional projections of the cells shown in (B) and (D), respectively. Scale bars, 2 μm.

To assess the influence of other DSB repair proteins on γH2AX-CD clustering, we introduced DSB tracks in Chinese hamster ovary cells deficient in one of two mechanistically distinct DSB repair pathways, homologous recombination and nonhomologous DNA end joining (8). Formation of both clustering (Subgroups Ia and Ib) and nonclustering (Group II) tracks was observed in these mutant cells, suggesting that the initial movement and adhesion of DSB-containing chromosome domains occurs upstream or independent of DSB repair.

Our results suggest why experiments using partial irradiation of cells with ultrasoft x-rays produced apparently immobile DSBs (5). Nuclei in those experiments absorbed a local dose of irradiation that was more than two orders of magnitude higher than that produced by an α-particle track. Every DSB resulting from the local x-irradiation would be embedded in a dense cloud of adhesive γH2AX-CDs that would prevent the DSB-containing chromosome domains from detectably moving. The use of lower doses delivered by local α irradiation reveals that the positions of DSB-induced γH2AX-CDs are not necessarily fixed and can move to cluster together. The clusters contain multiple γH2AX-CDs and multiple foci of the Mre11 and/or Rad51 DSB repair proteins, which supports the notion that distant DSBs can be juxtaposed. Gathering of multiple DSBs has also been observed in Saccharomyces cerevisiae cells (22). However, for yeast cells it is suggested that in S phase, multiple DSBs are recruited to repair centers containing a high concentration of repair proteins (22). In mammalian cells, juxtaposition of multiple DSBs occurs primarily in G1 phase, and the individual breaks already contain repair proteins. Juxtaposition of multiple DSBs, even though conserved among eukaryotes, seems potentially dangerous, because malfunctioning of repair could generate genomic rearrangements. Although chromosome translocations could also arise from the interaction of a single broken chromosome with a nondamaged chromosome (23), our demonstration that DSB-containing chromosome domains are mobile and can interact supports the breakage-first theory to explain the generation of translocations between two broken chromosomes.

Supporting Online Material

Materials and Methods

Table S1


References and Notes

View Abstract

Navigate This Article