Differential Repair Protein Recruitment at Sites of Clustered and Isolated DNA Double-Strand Breaks Produced by High-Energy Heavy Ions

DNA double-strand break (DSB) repair is crucial to maintain genomic stability. The fidelity of the repair depends on the complexity of the lesion, with clustered DSBs being more difficult to repair than isolated breaks. Using live cell imaging of heavy ion tracks produced at a high-energy particle accelerator we visualised simultaneously the recruitment of different proteins at individual sites of complex and simple DSBs in human cells. NBS1 and 53BP1 were recruited in a few seconds to complex DSBs, but in 40% of the isolated DSBs the recruitment was delayed approximately 5 min. Using base excision repair (BER) inhibitors we demonstrate that some simple DSBs are generated by enzymatic processing of base damage, while BER did not affect the complex DSBs. The results show that DSB processing and repair kinetics are dependent on the complexity of the breaks and can be different even for the same clastogenic agent.


Results
Differences in lesion complexity within individual HZE particle tracks impact on the timing of the DDR. To investigate the influence of simple and complex DNA lesions on the early DNA damage response (DDR) we made use of the different parts of the radiation tracks of HZE ions delivered by a particle accelerator in combination with real time live cell microscopy. For our study we used the established human osteosarcoma (U2OS) cell lines stably expressing the GFP-tagged DSB repair factors NBS1 and 53BP1 to follow the recruitment of these factors to the damage sites after irradiation. After irradiation with 1 GeV/u Fe-ions, we observed a fast recruitment of NBS1-GFP to the charged particle induced dense damage along the ion track. Protein accumulation at the track becomes clearly visible around 50 s after the ion traversal ( Fig. 1; for full time series see Supplemental Movie 1).
Interestingly, some isolated foci ( Fig. 1 arrow heads) occurring off-track showing a slightly delayed recruitment of NBS1 to DSBs could be monitored. These sparsely occurring off-track foci are likely to be generated by δ-electron track ends and represent isolated DSBs. For NBS1-GFP, the occurrence of radiation induced foci (RIF) in mock treated samples was a rare event under the applied imaging conditions with most cells showing no foci formation. This indicates that the analysed delayed occurring foci are induced by the application of IR and not light-induced during microscopy (Supplemental Fig. S3). Quantitative analysis of NBS1 intensities at both nuclear wide in-track and off-track DSBs yielded average recruitment curves ( Fig. 2A) similar to the ones described in literature for low-energy ions 19 . Whereas the slope of the recruitment is nearly identical for the averaged curves, a slight, but significant delay is evident for the off-track compared to the in-track foci ( Fig. 2A) resembling the optical observation. Fitting the recruitment curves to a delayed starting mono-exponential saturation curve yielded nearly identical time constants (96 ± 4 s vs. 97 ± 5 s) but a significant, around 50% longer delay time (61 ± 2 s vs 90 ± 3 s) (Supplemental Fig. S4) Theses results suggest a generally altered timing of the DNA damage response in dependency of lesion complexity.
Simple off-track DSBs are repaired faster compared to clustered in-track DSBs. In order to increase the yield of off-track foci, in the following experiment high-energy uranium ions were used (750 MeV/u; Figs. 2B and 3). In addition, to examine if the different radiation quality in the densely ionising part of the track and off-track effects the persistence of the NBS1 signals at the DNA break sites, we prolonged the post-irradiation observation time. Due to the much higher LET of the U-ions (2020 keV/µm in water) in comparison to the Fe-ions (150 keV/µm), not only more dense and complex damage along the track is generated, but also the deposited off-track dose and thus the probability for the formation of off-track DSBs can be largely increased without increasing the number of direct traversals (Supplemental Fig. S1). After irradiation with uranium, multiple off-track RIF become clearly evident in the nuclei ( Fig. 3A; for full time series see Supplemental Movie 1), thus facilitating further quantitative analysis. A high number of radiation-induced NBS1-foci were observed in nuclei not directly traversed by a uranium ion (Fig. 3A, lower nucleus). The cells irradiated with uranium ions reveal again a significantly delayed recruitment of NBS1-GFP to early appearing off-track foci compared to the track core in the nuclear wide analysis (Fig. 2B), similar to the curves obtained after irradiation with 1 GeV/u Fe ( Fig. 2A). The off-track foci show a more transient recruitment indicating less persisting NBS1 binding, whereas along the track the signal saturated and remained stable over the whole observation time of around 45 min (Fig. 2B). The lack of persistence of the off-track NBS1-GFP recruitment at early responding foci was confirmed using 350 MeV/u Ni (Supplemental Fig. S5 and Movie 2), indicating that the ion charge and LET do not have a major impact on the observed kinetic differences between core and penumbra DNA damage. Individual DSBs in single particle-irradiated nuclei reveal variable delays in repair factor recruitment.
Whereas hitherto the general description of the dynamic behaviour was based on global (nuclear wide) analysis of NBS1-GFP recruitment to radiation-induced damage sites, in the following we made use of the possibility to track individual foci in the live cell measurements. This allowed us to distinguish and address the recruitment of the damage sensing factor NBS1 to individual damage sites representing single DSBs for the off -track signal. Along the track core, which includes multiple (complex) DSBs in a single RIF or repair centre after high LET particle irradiation according to previous work from our group 20,21 and others 14,22 , a quite uniform fast and concerted recruitment to individual foci was observed. In contrast, individual off-track foci are characterised by an asynchronous response which is mainly governed by a distribution of individual delays (lag-phases; Fig. 3). Whereas less than 4% of NBS1-RIF inside the track showed a pronounced delay, around 40% of the observed off-track foci displayed lag phases larger than 2 min. The rising slope did not depend on the observed delay times for individual foci and no dependency of the delay to the radial distance from the track was observed for the off-track foci. Furthermore, around 50% of the off-track foci showed a clearly transient binding behaviour compared to <20% inside the track, reflecting the different lesion density and complexity. To rule out that the individually delayed damage response is not a unique feature of NBS1 we irradiated U2OS-cells expressing 53BP1, a mediator and important regulator in the DSB damage response known to form RIF in chromatin surrounding DSBs. After irradiation with 750 MeV/u uranium ions, 53BP1 accumulations along the tracks of traversing particles as well as multiple off-track foci in traversed and non-traversed nuclei could be detected ( Besides a slower recruitment (consistent with 53BP1 downstream role in the damage processing) and more pronounced delays, the DSB repair factor 53BP1 showed a similar behaviour as NBS1 regarding a more transient binding at off-track foci (with a maximum around 15-20 min post-irradiation) as well as the occurrence of individually delayed formation of off-track RIF (Fig. 4B,C). Again, quantitative analysis of lag-phases indicates a significant and clear difference between the relatively prompt recruitment of 53BP1 to the complex lesions inside the track and the broad distribution of delays for the off-track foci (Fig. 4D). For 53BP1, around 80% of off-track RIF show delay times >300 s after irradiation with uranium ions compared to only 22% in the track-core having lag phases larger than 300 s. More than 30% off track foci showed a lag phase exceeding 10 min. The 53BP1 kinetics emphasises the variability in the response to individual damage sites thus confirming the general observations obtained with NBS1.

X-ray induced DSBs are similar to simple off-track DSBs.
To corroborate if off-track foci represent less complex DNA lesions (isolated DSBs) like the ones induced by low-LET photon irradiation and to confirm the variability of the early DDR response at individual DSBs, we compared the live cell recruitment behaviour of NBS1 evaluated using the same microscope setup, but now coupled to a 35 kV X-ray source delivering relatively high dose rates (around 0.85 Gy/s). During X-ray irradiation we measured the recruitment of NBS1-GFP and 53BP1-GFP to radiation-induced RIF. Mean recruitment kinetics after X-rays match well to the ones observed for off-track foci using charged particles (Fig. 5).
Thus also photon-induced DSBs show a transient binding of NBS1, with a signal intensity peaking at 10-15 min. Moreover, the analysis of single X-ray induced foci revealed a broad range of individually different kinetic traces (Fig. 5C) with some clearly delayed recruiting DSBs. A fraction of ~35% of NBS1 foci were delayed www.nature.com/scientificreports www.nature.com/scientificreports/ by more than 2 min (Fig. 5D). Compared to NBS1-GFP, 53BP1-GFP showed larger, more clear and distinct RIF after X-ray irradiation, thus facilitating the evaluation (Supplemental Fig. S6A and Movie 4). Whereas mock irradiated samples develop around 1-2 foci/hour on average for 53BP1 under the applied imaging conditions, 0.85 Gy of X-rays induced around 28 foci in accordance with the expected number of DSBs 23,24 (Supplemental Fig. S3 and S6). Similarly to NBS1, the mean X-ray response of 53BP1 resembles the one obtained for off-track foci produced by HZE ions, supporting the similarity between X-and δ-rays (Supplemental Fig. S6B). In addition, 53BP1 recruitment to individual X-ray induced foci was characterised by a wide distribution of delays (Supplemental Fig. S6D) thus corroborating our previous results. In conclusion, isolated DSBs show a remarkable variability in their damage response suggesting a delayed conversion of non-DSB lesions into DSBs.
Delayed occurring DSBs are not due to DNA replication. In order to rule out that the observed delayed NBS1 recruitment is coupled to DSBs generated during replication (e.g. at stalled replication forks), we added a 5-ethynyl-2′-deoxyuridine (EDU)-pulse after irradiation and revisited the nuclei recorded during live cell imaging after fixation and staining of the EDU (S-Phase) and centromere protein-F (CENPF; G2-marker) (Supplemental Fig. S7A). Analysing cell cycle specific EDU and CENPF negative (G1) cells we obtained clear indication of delayed recruiting DSBs. This was evident in both G1-phase as well as S-phase (Supplemental Fig. S7C,D). In addition, the velocity of recruitment starting from the delay time (lag phase subtracted) was found to be nearly identical for G1 and S-phase cells (Supplemental Fig. S7B). In conclusion, as similar recruitment behaviour was observed in G1 phase and S-phase, we demonstrated that the observed delayed DSBs are not replication-dependent. (especially for NBS1) led us to the hypothesis that delayed foci might represent DSBs not generated directly by radiation, but rather through the repair-processing of complex non-DSB lesions. To address the role of damage conversion leading to delayed formation of DSBs, we addressed the influence of BER on the lag phase distribution of late occurring foci by inhibiting core factors. During the course of BER, damaged bases or abasic sites are converted into single-strand breaks (SSBs) by endonuclease activity. In case of clusters of non-DSB lesions or the existence of a SSB in near vicinity on the opposite DNA strand, this conversion may lead to a DSB 25 . The incision step is predominantly done by the apurinic endonuclease APE1, but strand cleavage of the sugar-phosphate backbone can also occur after OGG1 mediated removal of the damaged base via β-elimination at the resulting abasic site 26 . Therefore both reactions were chosen as targets for inhibition. Inhibition of APE1/OGG1 by a combination of methoxyamine hydrochloride and OGG1 inhibitor_O8 reduced the number of observed DSBs after X-rays by around 40%, indicating a major contribution of BER to DSB formation after low-LET radiation exposure. Furthermore, BER inhibition resulted in longer NBS1 foci delays, demonstrating the influence of BER on the average timing of DSB induction and recognition (Fig. 6).
Nevertheless, the observed shift to longer delay times is most probably mainly governed by the loss of early occurring DSBs upon inhibition, indicating that the BER incision is a fast process. In fact, BER is not substantially contributing to the formation of delayed DSBs associated to the prolonged lag-phases of NBS1 recruitment. Rather other mechanisms like conversion of radiation-induced labile sites by post-irradiative chemical modification of sugar lesions 27,28 may contribute to the conversion of clustered non-DSB lesions. The shift in the lag phase distribution to longer delay times was confirmed by single treatments with the APE1 inhibitor methoxyamine as well as a second APE1 inhibitor (CRT0044876) targeting the nuclease activity of APE1 or by www.nature.com/scientificreports www.nature.com/scientificreports/ oxoguanine glycosylase inhibition (OGG1 inhibitor_O8). All single treatments led to similar delays in the lag phase distribution of detected DSBs, thus supporting the hypothesis of conversion of clustered base damage into DSBs (Supplemental Fig. S8B-D). However, in contrast to the combination of APE1 and OGG1 inhibitors, single inhibitor treatments did not lead to a significant reduction of radiation-induced number of DSBs. Despite a delayed recognition of DSBs, inhibiting APE1 had no major impact on cellular clonogenic survival, indicating that the delay of DSB detection does not lead to enhanced radiosensitivity and cell death (Supplemental Fig. S8A).

The BER factor XRCC1 accumulates fast but transiently only on clustered in-track lesions.
To further elucidate the contribution of BER in the DDR after high-LET particle irradiation, we followed the real-time recruitment of the repair factor XRCC1 using human fibrosarcoma cells (HT1080) stably expressing EGFP-tagged XRCC1. After irradiation with 1 GeV/u Fe or 750 MeV/u U-ions, a clear, very fast and locally highly confined recruitment to the DNA damage along the trajectories could be observed without significant delay (see Supplemental Movies 5 and 6) similar to the one we described previously in mouse cells 29 . However, besides some weak changes in graininess, no clear off-track foci could be visualised, most probably due to the abundance of base damages and SSBs and the low number of accumulated molecules per lesion (Supplemental Fig. S9,A) complicating the analysis of recruitment to off-track damage. In contrast to the more stable binding of the DSB factors NBS1 and 53BP1, the recruitment of XRCC1 to the massive complex damage inside the track shows a transient accumulation peaking around 10 min after irradiation and subsequently decaying with a half-life of around 10 min (Supplemental Fig. S9B). This clearly demonstrates that not all repair factors become trapped at the massive damage induced by the highly ionising core of the HZE tracks. The transient nature of binding of XRCC1 points either to a displacement of XRCC1 by the DSB repair machinery at damage clusters or to a fast processing of the majority of SSBs and base damage even in case of clustered lesions.

Inhibition of BER does not affect NBS1 recruitment to clustered in-track DSBs. To address, if
BER inhibition generally slows down the recruitment of NBS1 to DSBs, we analysed the recruitment kinetics at the clustered in-track DSBs. Whereas inhibition of the incision step of BER clearly affected the recruitment to the isolated DSBs induced after X-rays, no change in the slope or distribution of delay-times was obtained for in-track RIF after irradiation with 350 MeV/u Fe-ions (Fig. 7). This indicates that the recruitment to the complex damage in the track is mainly governed by the high number of promptly induced DSBs and that the application of the BER inhibitors is not changing the general kinetics of NBS1.

Discussion
In this study we measured simultaneously real time recruitment kinetics of early DDR factors to clustered and simple DSBs in the same cell upon traversal by high-LET charged particles. Simple, off-track DSBs produced by δ-electrons of these particles have been predicted using biophysical modelling 30 , but hitherto have been studied only in fixed samples [31][32][33] , unavoidably missing the early activation of repair factors and the dynamics of individual foci. The complex, clustered DSBs formed along the core of the traversing HZE ion showed a prompt accumulation of the early DSB binding factor NBS1, in line with our previous measurements with very densely ionising low energy particles characterised by fast kinetics of NBS1 at the tracks due to fast ATM activation 19 . For both types of high-LET particles local doses in the range of kGy are deposited along the core leading to the generation of complex clustered DSBs with multiple DSBs in each focus 20,31 . The analysis of NBS1 accumulation at individual simple, off-track DSBs revealed a significant fraction of lesions showing a delayed onset of recruitment as well as a more transient binding to the damage. This observation was rather independent from the applied off-track dose www.nature.com/scientificreports www.nature.com/scientificreports/ as revealed by the usage of different HZE particles. Of note, real time measurements of repair factors, even those acting early in the DSB response, might not reflect the exact timing of DSB induction and repair in case some pre-processing at the damage site was needed. However, even larger scale chromatin remodelling after particle irradiation in heterochromatin has been shown to occur in the order of tens of seconds and will thus not delay the recruitment of early repair factors to a larger degree 29,34 . These general observations were confirmed using charged particles of varying atomic number and energy as well as 53BP1-RIF formation, a surrogate marker of DSBs 18 like γH2AX. Compared to γH2AX or 53BP1, NBS1 requires less pre-processing steps before the onset of recruitment. Thus the observation of partially delayed recruitment of both NBS1 and 53BP1 to individual DSBs corroborates our interpretation of an individually delayed DSB formation rather than local differences in protein activation. The shorter persistence of the protein accumulation at the simple, δ-ray-induced off-track DSBs reflects a faster processing and repair compared to the clustered DSB induced along the track core. This interpretation is consistent with experiments showing impaired or delayed repair at complex lesions 14,21,29,32,33,35,36 or increased accretion of 53BP1 to DSB-clusters versus single-DSBs 37 . Stalled repair clusters may persist for a substantial time in cells and lead to chromosomal breakage 11 and enhanced mutagenesis 38 .
The individually delayed recruitment of the off-track RIF produced by electrons emitted by fast heavy ions was similarly observed in RIFs produced by low-LET X-rays (Fig. 5 und Supplemental Fig. S6), indicating that both radiation types induce similar damage spectra including non-DSB clustered lesions. This is expected from the physical process of energy deposition by electrons produced by either a photon (X-rays) or an HZE ion (δ-rays) atomic ionisation 30 . In line with that notation, we have found no correlation between the lag times and the radial distance from the core for the off-track DSBs, suggesting that most δ-ray and X-ray induced DSBs are generated by the high energy deposition of the electrons toward the end of their range. Protein accumulation and release from RIF are similar for X-and δ-rays (Fig. 5), corroborating the evidence that the complexity of DSBs determines their repair kinetics. Compared to the off-track areas or after application of sparsely ionising radiation, where DSBs are produced mainly by indirect, radical-mediated action, a higher fraction of direct damage is induced inside the densely ionising ion tracks 39,40 . Nevertheless, these differences will not impact on the observed delay times as the timescale of the physico-chemical propagation of the highly reactive radicals are much shorter (µs) than the recruitment kinetics addressed in this study.
The contribution of BER-processing to the formation of DSBs after irradiation was demonstrated here by preventing the incision step of the DNA backbone via inhibition of the endonuclease activity, thus limiting the generation of SSBs. It has been shown that two opposite base lesions >3 bp apart or combinations of SSBs and base damage/loss results in DSB formation during the BER excision/ incision step 27,41,42 . The reduction in the number of early recruiting RIFs observed in our study is in agreement with a fast operation of BER incision on lesions containing 2 AP sites as bi-stranded cluster 38 or other combinations of clustered non-DSB lesions which can be efficiently converted into DSBs 41 . Our results suggest that besides participation of BER in DSB production, this pathway is not responsible for the observed delayed recruiting fraction of DSBs. The inhibition of the early www.nature.com/scientificreports www.nature.com/scientificreports/ steps of BER had no significant impact on the recruitment of NBS1 to in-track RIF, indicating that the severe clustered damage in the high-LET core is dominated by prompt DSBs, which already activate the local chromatin repair response independently of BER.
Delayed NBS1 recruitment was observed throughout the cell cycle in interphase cells (Supplemental Fig. S6), thus ruling out replication-dependent conversion of non-prompt DSBs 43 as a major source for late responding RIFs. We therefore propose that other clusters of radiation-induced labile non-DSB lesions are responsible for our observation. Heat-labile sites have been described as source of DSBs analysed using pulse-field gel electrophoresis 27 , but it has been shown that a subset of these can be converted into DSB also under physiological conditions in cells, thus leading to the generation of delayed occurring DSBs 27 .
In conclusion, our study shows that DDR kinetics is strongly influenced by damage complexity. While additional factors like the local chromatin environment (e.g. heterochromatin 44 ) can contribute to differences as well, we propose that enzymatic processing and conversion of non-DSB lesions play major roles in the repair kinetics of simple DSBs. Damage induced by isolated electrons is processed with similar mechanisms, independently of their source (photons or charged particles). Recruitment of DNA repair proteins is faster to clustered, complex DNA lesions and dominated by prompt DSB formation which can be repaired without involvement of BER. The current mechanistic models of radiation action, aiming to link DNA damage and repair to late cellular effects [45][46][47][48][49][50] , currently use mostly protein recruitment kinetics from laser-induced DNA lesion experiments 51,52 or the in-track RIF produced by heavy ions 19,51,53 , thus neglecting the damage induced by the δ-rays, which transport roughly 50% of the energy deposited by the primary ion 54 . Our results provide the necessary data to treat separately simple and clustered DNA in biophysical modelling of radiation action.
Cell irradiation. For ion irradiation 70 000 cells/mL were seeded 24 h prior to experiments on 4-well chamber slides (Thermo Scientific), for X-Ray irradiation 75 000/mL cells on petri dishes with glass bottom (35 mm Ø; 10 mm thickness; Greiner Bio-One GmbH). In the experiment performing retrospective cell cycle analysis after irradiation, cells were seeded on petri dishes with a grid-500 etched glass bottom (35 mm Ø; 0,17 mm thickness; Ibidi, Germany) to be able to revisit cells after fixation. Charged particle irradiation was performed using a horizontal beam (fluence 5·10 6 cm −2 ) at the heavy-ion synchrotron (SIS) of the GSI Helmholtz Centre for Heavy Ion Research using different HZE species 56 Fe (1 GeV/u; LET 150 keV/µm, 1.2 Gy or 350 MeV/u; LET 220 keV/ µm, 1.8 Gy), 238 U (750 Me V/u; LET 2020 keV/µm, 16 Gy), 59 Ni (350 MeV/u; LET 255 keV/u, 2.0 Gy) as indicated. Roughly a fraction of half of the dose can be considered to be delivered to the track centre and penumbra, respectively. Cell irradiation with X-Ray was performed as described before 57,58 . Briefly, a 35 kV X-ray tube (GE Inspection Technology, Ahrensburg, Germany) which was operated at 80 mA delivering a dose rate of 0.85 Gy/s at the cell layer was used. Dosimetry was done using a Unidos dosemeter equipped with a soft X-ray ionisation chamber TM23342A (PTW, Freiburg, Germany) applying a correction factor (1.2) for the dose increase at the glass surface. For inhibition of BER pathway methoxyamine hydrochloride (20-35 mM) 59 , OGG1 inhibitor_O8 (50-75 µM) 60 and CRT0044876 (150-240 µM) 61 (all purchased from Sigma-Aldrich) were incubated with cells 1-2 h prior to irradiation as indicated.
Live cell imaging and data analysis. Live cell observations were performed using a modified remote controlled OLIMPUS IX71 or IX73 microscope for both ion and X-ray irradiation using an UPlanFL60x/1.2 water or 100 × /1.4 oil immersion lens and a 1.6 x optovar as described in 19 . Except for the experiment using the 350 MeV/u Fe beam, where an Andor Zyla 4.2 sCMOS camera was used, image acquisition was done with an EM-CCD camera type DU-888 (Andor Technology, Belfast, Ireland). For time lapse imaging of z-stacks, AndorIQ software was used. Fluorescence was excited with the CoolLED system or monochromator Polychrome V (TILL Photonics GmbH, Gräfelfing, Germany), respectively. Acquired z-stacks (0.5 μm steps) were deconvoluted using the Huygens essential software (SVI, Netherlands). Image analysis of nuclear wide recruitment kinetics was performed with the software package ImageJ (https://imagej.nih.gov/ij/; NIH, USA) as described in 19 . To facilitate single focus analysis, a software package (ImageD -Version 1_6-8 with plugin Foci Tracking) was developed in house. Here, foci were automatically detected in maximum projections of the z-stacks using locally adaptive thresholds with the option of manual corrections. The automatic evaluation of lag phases is based on the variation in the standard deviation between the selected time windows with relative fluorescence intensity of the lag time ≥10% of the maximal intensity value of the concerned focus as an additional parameter. The measurements were double normalised to the signal loss during the image acquisition and to the plateau, according to 19