53BP1 promotes NHEJ of telomeres by increasing chromatin mobility

Abstract

Double strand breaks (DSBs) activate the ATM kinase which promotes the accumulation of DNA damage factors in the chromatin surrounding the break. The functional significance of the resulting DNA damage foci is poorly understood. Here we show that 53BP1, a component of DNA damage foci, changes the dynamic behavior of chromatin to promote DNA repair. We used conditional deletion of the shelterin component TRF2 from mouse cells to deprotect telomeres, which, like DSBs, activate the ATM kinase, accumulate 53BP1, and are processed by non-homologous end-joining (NHEJ)^1,2. Deletion of TRF2 from 53BP1-deficient cells established that NHEJ of dysfunctional telomeres is strongly dependent on the binding of 53BP1 to damaged chromosome ends. To address the mechanism by which 53BP1 promotes NHEJ, we used time-lapse microscopy to measure telomere dynamics before and after their deprotection. Imaging showed that deprotected telomeres are more mobile and sample larger territories within the nucleus. This change in chromatin dynamics was dependent on 53BP1 and ATM but did not require a functional NHEJ pathway. We propose that the binding of 53BP1 near DNA breaks changes the dynamic behavior of the local chromatin, thereby facilitating NHEJ repair reactions that involve distant sites, including joining of dysfunctional telomeres and AID-induced breaks in immunoglobulin Class-Switch Recombination (CSR).

Previous work has shown that mouse telomeres lacking TRF2 are processed by a Ku70/DNA ligase IV-dependent NHEJ reaction^1,2 that requires ATM kinase signaling and is stimulated by the ATM targets γ−H2AX and MDC1^3,4. Here we focus on 53BP1, a third ATM target, which accumulates at DSBs and deprotected telomeres^5–8. The interaction of 53BP1 with chromatin involves the binding of its Tudor domains to H4-K20diMe and an MDC1-dependent interaction with γ-H2AX^9–14. While 53BP1 is not strictly required for DNA damage signaling, homology directed repair (HDR), or NHEJ in the context of V(D)J recombination, NHEJ of DSBs in CSR is severely affected by 53BP1 deficiency^15,16. In the absence of 53BP1, DSBs in different switch regions fail to join successfully, resulting in a predominance of intra-switch recombination events¹⁷. It has been proposed that 53BP1 might either facilitate synapsis of DNA ends^17,18 or ‘shepherd’ NHEJ factors to the break¹⁹.

We generated SV40-LT immortalized TRF2^F/−53BP1^−/− and TRF2^F/−53BP1^+/− MEFs^1,20 (Suppl. Fig. 1a, b) and assayed the frequency of telomere fusions in metaphase spreads collected 120 hours after Cre-mediated deletion of TRF2 (Fig. 1a). Whereas 53BP1-proficient cells showed the expected level of telomere fusions (33% of telomeres fused after 4 PD), the rate of NHEJ in 53BP1^−/− cells was at least 50-fold lower, nearly as low as in DNA ligase IV deficient cells^1,2. The effect of 53BP1 on telomere fusions was also obvious when assayed by in-gel hybridization assay (Fig. 1b). 53BP1-deficient cells failed to accumulate high molecular weight fusion products after deletion of TRF2 and showed no loss of the telomeric 3’ overhang, a second index for telomere fusion (Fig. 1b, c). In fact, the overhang signal increased, similar to what is observed when TRF2 is deleted from Ku70- or DNA Ligase IV-deficient cells^1,2. In contrast to NHEJ, DNA damage signaling was not affected by the absence of 53BP1 (Fig. 1d, e). 53BP1 deficiency affected neither the phosphorylation of ATM and its target Chk2 nor the presence of γ-H2AX, MDC1, and NBS1 at dysfunctional telomeres (Fig. 1d, e, Suppl. Fig. 2a, b). The NHEJ defect was also not due to a change in cell cycle progression. After deletion of TRF2, the cells underwent the same number of cell divisions regardless of their 53BP1 status (Fig. 1f) and their S-phase index was not affected by 53BP1 deficiency (Suppl. Fig. 1c).

Figure 1. Requirement for 53BP1 in NHEJ of dysfunctional telomeres, but not DNA damage signaling.

a, (Top) Metaphase chromosomes after deletion of TRF2 from the indicated cells. Telomeric FISH, green; DAPI, red. (Bottom) Summary of the effect of 53BP1 on telomere fusions at 120 hours after TRF2 deletion.

b, In-gel assay for the 3’ overhang (left) and total telomeric DNA (right) after deletion of TRF2 from 53BP1-proficient and -deficient cells.

c, Quantification of overhang signals in (b) (mean±SD; n=3).

d, IF for presence of γ-H2AX at telomeres after deletion of TRF2 from 53BP1-proficent and -deficient cells.

e, Immunoblots for phosphorylated ATM and Chk2 after TRF2 deletion.

f, Proliferation of MEFs of the indicated genotype and treatment (mean±SD; n=3).

We next tested whether the Tudor domain-mediated recognition of H4-K20diMe contributed to the promotion of NHEJ. TRF2^F/−53BP1^−/− cells were complemented with wild type human 53BP1 or a Tudor domain mutant (53BP1-D1521A) with impaired binding to H4-K20diMe¹⁴ (Fig. 2a). Although the two forms of 53BP1 were expressed at levels comparable to endogenous 53BP1 (Fig. 2b and data not shown), the recruitment of 53BP1-D1521A to deprotected telomeres was significantly diminished (Suppl. Fig. 3). In contrast to wild type 53BP1, which reconstituted the NHEJ of dysfunctional telomeres, cells expressing the Tudor mutant of 53BP1 showed a considerable delay in the fusion of telomeres lacking TRF2 (Fig. 2c, d).

Figure 2. Optimal NHEJ of dysfunctional telomeres requires interaction of 53BP1 with H4-K20diMe.

a, Schematic of the domain structure of 53BP1.

b, Immunoblot for 53BP1 expression in TRF2^F/−53BP1^−/− cells complemented with empty vector, wild type (WT), or the Tudor mutant 53BP1-D1521A.

c, Telomere fusions in metaphase spreads of TRF2^F/−53BP1^−/− MEFs complemented as indicated and treated with Cre as specified.

d, Quantification of telomere fusions in (c) (mean±SD; n=3).

The finding that the Tudor domain mutation in 53BP1 did not abrogate NHEJ suggested that a second, H4-K20diMe-independent, interaction of 53BP1 with chromatin contributes to NHEJ. As MDC1 is required for the stable association of 53BP1 with DSBs^9–14, we generated TRF2^F/FMDC1^−/− MEFs¹⁸ and determined their ability to execute the telomere fusion reaction. As expected, the binding of 53BP1 near deprotected telomeres was impaired by MDC1 deficiency, despite normal levels of H2AX phosphorylation (Suppl. Fig. 4a, b). MDC1 deficiency resulted in a delay in the NHEJ of telomeres but did not abrogate this process (Suppl. Fig. 5a, b), confirming our previous findings with MDC1 shRNAs³. Furthermore, inhibition of MDC1 with shRNAs affected the residual NHEJ of telomeres in the context of the 53BP1 Tudor domain mutant (Suppl. Fig. 5c–e).

These data establish an important role for chromatin-bound 53BP1 in the NHEJ of deprotected telomeres. Previous data showed that 53BP1 contributes to NHEJ in CSR^15,16 but not in other settings such as V(D)J recombination. We argued that the crucial difference between the 53BP1-dependent and –independent NHEJ reactions might be the distance between the DNA ends involved. Whereas the two ends generated by RAG1/2 or chromosome-internal DNA damage are close together, the DNA ends generated by AID in CSR are often far apart as are dysfunctional telomeres, which are processed by NHEJ in G1²¹ when chromosome ends are dispersed throughout the nucleus.

We turned to time-lapse microscopy to address whether 53BP1 altered the synapsis and/or dynamics of deprotected telomeres. To image telomeres, we used an EGFP-tagged version of the shelterin component TRF1, which remains associated with telomeres when TRF2 is removed (Fig. 3a–c, Suppl. Fig. 6). Overexpression of this and other forms of TRF1 does not affect the protective function of telomeres²² (data not shown). To mark sites of DNA damage signaling, we fused mCherry to aa 1220–1711 of 53BP1 (mCherry-BP1-2, see Fig. 2a and Fig. 3a, b), creating a fusion protein that lacks most of the functional domains of 53BP1. Upon deletion of TRF2, mCherry-BP1-2 re-localized to telomeric sites containing EGFP-TRF1 (Fig. 3c). The mCherry-BP1-2 allele did not affect the frequency of telomere fusions in 53BP1-proficient cells nor did it restore the NHEJ defect in 53BP1-deficient cells (Suppl. Fig. 7a), establishing that it is a neutral marker for DNA damage in this context. Furthermore, key imaging experiments repeated in cells lacking mCherry-BP1-2 gave the same outcome (Suppl. Fig. 7b,c and see below).

Figure 3. 53BP1 promotes mobility of dysfunctional telomeres.

a, Schematic of the live-cell imaging experiments.

b, Immunoblotting for EGFP-TRF1, mCherry-BP1-2, and TRF2.

c, Co-localization of EGFP-TRF1 and mCherry-BP1-2 fluorescence signals.

d, Representative traces of telomeres in indicated MEFs expressing EGFP-TRF1 and mCherry-BP1-2. Snapshots taken at indicated timepoints.

e, 10 representative telomeres from indicated MEFs, treated as specified, were tracked as in (d); cumulative distance at each time-point was plotted against time.

f, Frequency distribution of the cumulative distance traveled in one representative experiment in MEFs of the indicated genotype and treatment (median±SD.; n=55; p calculated using a two-tailed Mann-Whitney test).

g, Frequency distribution of the maximum displacement from the starting point registered by individual telomeres during the imaging session described in (f). (median±SD and p-values calculated as in (f)).

h, i, Analysis as in (f) in MEFs of the indicated genotypes and treatments, expressing EGFP-TRF1 marker only (n=50).

To analyze the effect of 53BP1 on the movement of dysfunctional telomeres, time-lapse microscopy was performed at an early stage after deletion of TRF2 (72–84 hr post-Cre) when most TRF1-marked sites still represent free chromosome ends³. Initial analysis of EGFP and mCherry signals indicated that dysfunctional telomeres became more mobile but only when 53BP1 was present (Suppl. Video 1 and 2). We observed occasional apparent fusion events in 53BP1-proficient cells (Suppl. Video 3) but no potential fusions were observed in 53BP1-deficient cells even for telomeres that were closely apposed (Suppl. Video 4).

To obtain a quantitative measure of telomere mobility, individual EGFP-TRF1/mCherry-BP1-2-labeled dysfunctional telomeres were traced in deconvolved and projected images (Fig. 3a, d; Suppl. Fig. 8a, b). In parallel, EGFP-TRF1-marked functional telomeres were analyzed in cells not treated with Cre. Functional telomeres in 53BP1-proficient and -deficient cells traveled in a constrained random walk over a total path of approximately 3.7–3.9±0.7 µm at a median speed of 180–190 nm/min⁻¹, which is comparable to movement of human telomeres (Fig. 3e, f; Suppl. Figs. 7b, 9a)²³. In contrast, dysfunctional telomeres were significantly more mobile, traveling at a speed of 270–360 nm/min over a cumulative distance ranging from 5.4 to 7.2 µm (Fig. 3e, f; Suppl. Figs. 7b, 9a). The mobility increase associated with telomere dysfunction was significantly attenuated in cells lacking 53BP1 (Fig. 3e), resulting in a median cumulative distance traveled of 4.4±0.2 µm. The presence of the mCherry-BP1-2 DNA damage marker did not affect the outcome; the same results were obtained after deletion of TRF2 from 53BP1^+/− or 53BP1^−/− cells that only contained the EGFP-TRF1 marker (Suppl. Fig 7b, c). Importantly, in both settings the dysfunctional telomeres in 53BP1-proficient cells sampled larger territories than functional telomeres (maximal displacement 1.2±0.3 µm versus 0.51±0.29 µm; Fig. 3g, Suppl. Figs. 7c, 9b). Assuming that telomeres sample spheres within the nucleus, a 2-fold change in the radius of the 2D territory will increase the volume sampled by a dysfunctional telomere by a factor 8. Furthermore, a substantial fraction (> 10%) of the dysfunctional telomeres roamed well beyond 2 µm whereas none of the 115 functional telomeres analyzed moved beyond that distance (Fig. 3g, Suppl. Fig. 7c). When 53BP1 was not present, the median maximal displacement of the dysfunctional telomeres was only 0.8±0.2 µm, indicating movement within a significantly smaller territory and only one out of 115 dysfunctional telomeres sampled an area beyond 2 µm (Fig. 3g, Suppl. Fig. 7c). These data establish that telomeres become more mobile and sample larger territories when they are deprived of their normal protection and that this change in their dynamic behavior is promoted by 53BP1. Assuming that the rate of NHEJ correlates with the probability of an encounter between two telomeres, the ability of 53BP1 to expand the 3 dimensional space visited by dysfunctional telomeres is a likely explanation for its effect on telomere fusion. We confirmed that the absence of the NHEJ reaction itself could not account for the slower movement as telomeres in DNA ligase IV-deficient cells still showed a considerable increase in their mobility upon TRF2 deletion (Fig. 3h, Suppl. Fig. 9c, d).

The ability of 53BP1 to promote the mobility of dysfunctional telomeres could explain why ATM kinase, its upstream regulator, is required for telomere fusions⁴. Indeed, similar to 53BP1-deficient cells, dysfunctional telomeres failed to gain their maximal mobility in ATM-deficient cells (Fig. 3i, Suppl. Fig. 9c, d).

Next, we tested whether 53BP1 led to global chromatin mobility in response to DNA damage. We tracked functional telomeres in wild type or Lig4^−/− cells immediately after treatment with γ-irradiation or after allowing the cells to recover. In all cases, we found that the mobility of the telomeres was unaffected by the induction of damage elsewhere in the genome (Suppl. Fig. 10) arguing that increased mobility is a local event, taking place at the site of damage where 53BP1 accumulates.

As 53BP1 is the first DNA damage factor to be implicated in the movement of sites of DNA damage, it will be important to determine the mechanism by which it affects chromatin mobility. Latrunculin A did not alter the gain in mobility of deprotected telomeres (Suppl. Fig. 11a), arguing against an actin-driven process. Similarly, the HDAC inhibitor trichostatin A had no effect on the mobility of the deprotected telomeres, making it unlikely that 53BP1 acts through its interaction with HDAC4²⁴ (Suppl. Fig. 11b). We imagine that 53BP1 dislodges a factor that restricts chromatin mobility or that 53BP1 endows the chromatin with dynamic features, for instance by altering higher order interactions that affect the thermal persistence length.

These data reveal that the DNA damage factor 53BP1 is crucial for the NHEJ reaction at dysfunctional telomeres and suggest a mechanism by which 53BP1 promotes telomere fusions. Telomeres, like other chromosomal sites, have a limited range of motion, showing constrained diffusion within a territory with a radius of ≤0.5 µm (this work and refs. 23 and 25^23,25). When telomeres become damaged, their mobility increases significantly, which is expected to increase the chance that two dysfunctional telomeres become closely apposed, thereby allowing the NHEJ machinery to engage the two DNA ends. Given that 53BP1 is also important for CSR, we suggest that the change in chromatin dynamics also promotes the joining of AID-induced breaks. In an accompanying paper by Nussenzweig and colleagues, this paradigm is further extended to distant DSBs in V(D)J recombination. We therefore suggest that 53BP1 is a facilitator of NHEJ at all distant DNA ends, acting by increasing the mobility of the local chromatin. As joining of distant DNA ends promotes chromosome rearrangements, we consider it unlikely that the main role of 53BP1 is to promote NHEJ at DNA ends that are far apart. It appears more likely that the increased chromatin mobility we observe reflects an attribute of 53BP1-containing chromatin that is important for DNA repair at all DSBs, including those that are closely apposed and breaks that are processed by HDR. We imagine that at telomeres, this chromatin change leads to increased mobility, whereas at chromosome-internal DSBs the same change in the chromatin may facilitate their repair in other ways.

Methods Summary

MEFs from E13.5 embryos obtained from crosses between TRF2^F/− and 53BP1^+/− mice^1,20 were isolated and immortalized at passage 2 with pBabeSV40LT (a gift from G. Hannon). Cre was introduced by retroviral infection with pMMP Hit&Run Cre²⁶. For time-lapse microscopy, TRF2^F/−53BP1^+/− and TRF2^F/−53BP1^−/− cells, expressing EGFP-TRF1 and mCherry-BP1-2; or TRF2^F/−Lig4^−/−p53^−/−1 and TRF2^F/−ATM^−/−4 cells expressing EGFP-TRF1 only, were imaged untreated or 72 h after Cre-mediated deletion of TRF2. Time-lapse images were acquired using a DeltaVision RT microscope system (Applied Precision) in 3D in both EGFP and mCherry channels using SoftWoRx software every 30 seconds over 20 minutes. Images were deconvolved, projected in two dimensions and tracking analysis was performed with ImageJ. Data were collected from 3 independent experiments (n≥50) and Mann-Whitney statistical analysis was performed using Prism software.

Full methods accompany this paper.