Replicated chromatin curtails 53BP1 recruitment in BRCA1-proficient and -deficient cells

DNA double-strand breaks can be repaired by two competing mechanisms, non-homologous end-joining (NHEJ) and homologous recombination (HR). Whether one or the other repair pathway is favored depends on the availability of an undamaged template DNA that allows for homology-directed repair. The tumor suppressor proteins 53BP1 and BRCA1 are considered antagonistic players in this repair pathway choice, as 53BP1 restrains DNA end resection, whereas BRCA1, together with its partner protein BARD1, displaces 53BP1 from damaged replicated chromatin and promotes HR. How cells switch from a 53BP1-dominated to a BRCA1-dominated response as they progress through the cell cycle is incompletely understood. Here we reveal, using high-throughput microscopy and applying single cell normalization to control for increased genome size as cells replicate their DNA, that 53BP1 recruitment to damaged replicated chromatin is inefficient in both BRCA1-proficient and BRCA1-deficient cells, in comparison to 53BP1 accumulation at damaged unreplicated chromatin. These findings substantiate a dual switch model from a 53BP1-dominated response in unreplicated chromatin to a BRCA1-BARD1-dominated response in replicated chromatin, in which replication-coupled dilution of 53BP1’s binding mark H4K20me2 functionally cooperates with BRCA1-BARD1-mediated suppression of 53BP1 binding. More generally, we suggest that appropriate normalization of single cell data, e.g. to DNA content, provides additional layers of information, which can be critical for quantifying and interpreting cellular phenotypes.


Introduction
The balance between non-homologous end-joining (NHEJ) and homologous recombination (HR) has important implications for maintenance of genome stability, for exploiting DNA damage response (DDR) defects as vulnerabilities in cancer therapy, and for harnessing the full potential of CRISPR/Cas9-mediated genome editing and gene therapy [1][2][3] . The choice between NHEJ and HR is closely linked to the cell cycle and to whether or not an undamaged homologous stretch of DNA is available as a template for HR [4][5][6] . While 53BP1-RIF1-Shieldin restrains DNA end resection of DNA double-strand breaks (DSBs) in the absence of a homologous template strand, after DNA replication BRCA1-BARD1 counteracts 53BP1 binding to damaged chromatin and promotes HR reactions [7][8][9][10][11][12] . The antagonism between 53BP1-RIF1-Shieldin and BRCA1-BARD1 has important implications for cancer therapy, in particular for targeting BRCA-deficient tumors with PARP inhibitors, and for improving CRISPR/Cas9-mediated gene editing.
HR is confined to the S and G2 phase of the cell cycle, yet how the DDR discriminates between replicated and unreplicated areas of the genome during S phase progression has only started to emerge recently. Such discrimination is important for HR reactions to be favored when DSBs occur in replicated DNA, and mutagenic end resection to be prevented when DSBs occur in areas of the genome that were not replicated yet.
Accumulating evidence suggests that this discrimination is linked to the different chromatin makeup ahead of and behind replication forks [10][11][12][13][14] . Of particular interest for the antagonism between the 53BP1 and BRCA1 protein complexes is the H4K20 dimethylation mark (H4K20me2), which is abundant in unreplicated chromatin on parental histones and absent from newly incorporated histones. Hence H4K20me2 is diluted in nascent replicated chromatin and only restored in mature chromatin in late G2/M [10][11][12][13][14] . 53BP1 binds H4K20me2 via its tandem tudor domain (TTD), and both H4K20me2 and the TTD are required for 53BP1 recruitment to DSBs 15,16 . Conversely, unmethylated H4K20 (H4K20me0) in replicated chromatin is recognized by the HR-promoting protein complexes TONSL-MMS22L and BRCA1-BARD1 12,14 .
Among the evolving methods to probe cellular stress responses at the single cell level is quantitative image-based cytometry (QIBC), which uses automated high-content microscopy in a cytometry-like fashion to stage individual cells according to their position in the cell cycle [17][18][19] . Besides quantification of multiple cellular parameters in large cell populations, when such microscopy-based single cell measurements are accurate enough to discriminate cells according to their position in the cell cycle, they allow for cell cyclerelated data normalization, e.g. to the size of a cell or its nucleus, or to the replication status of the genome. As multiple cellular functions are tightly linked to cell cycle progression and to the associated duplication of cellular contents, including duplication of the genome and related entities, we suggest, using the example of cell cycle regulated 53BP1 recruitment to damaged chromatin as paradigm, that normalization of cell cyclerelated phenotypes to DNA content and thus cell cycle position, when implemented in quantitative cell image analysis pipelines, can reveal additional layers of information that facilitate biological interpretation of quantitative cell imaging data.
It was recently established that the ankyrin repeat domain (ARD) of BARD1 binds to H4K20me0 and brings the BRCA1-BARD1 complex to DNA breaks when a sister chromatid is available for DSB repair by HR, thus illuminating the mechanistic underpinnings of BRCA1-mediated displacement of 53BP1 from replicated chromatin 12 . An unresolved question is, however, whether the antagonism between BRCA1-BARD1 and 53BP1 alone is sufficient to explain the displacement of 53BP1 from damaged replicated chromatin, or whether the dilution of H4K20me2 behind replication forks directly impacts 53BP1 recruitment and thereby affects its functions even when BRCA1-BARD1 functions are impaired. Elucidating the relative contribution of BRCA1-BARD1-related versus unrelated factors for antagonizing 53BP1's functions could provide critical information on DSB repair pathway choice with implications for utilizing HR and NHEJ in the context of genome editing and for targeting the DDR in cancer therapy.

Results and discussion
By means of software-assisted image segmentation and feature extraction, QIBC allows measurements of cellular phenotypes in large, asynchronously growing cell populations. The delta in efficiency of 53BP1 recruitment in gated cell subpopulation averages based on DNA content (2N versus 4N) was even more pronounced when G1 cells were compared to cells in late S / early G2 based on DAPI and Cyclin A staining, both in BRCA1-BARD1-proficient and -deficient cells (Figure 5a-d). Similarly, when H4K20me2 was used to define G1 versus late S / early G2 cells (i.e. excluding cells in mid and late G2 with gradually restored H4K20me2), the difference in 53BP1 recruitment was more pronounced compared to cells gated merely based on their DNA content (Figure 6a-d).
Consistent results were obtained with the breast cancer cell line SUM149PT carrying the BRCA1 2288delT mutation and allelic BRCA1 loss (Extended Data Figure 5a-c). Taken together, we conclude that inefficient 53BP1 recruitment to damaged replicated chromatin is an inherent feature that occurs both in BRCA1-deficient and, in a more pronounced manner, in BRCA1-proficient cells, and we therefore suggest that replication-coupled dilution of H4K20me2, in addition to enabling H4K20me0-mediated BRCA1-BARD1 recruitment, also directly affects the efficiency of 53BP1 recruitment in response to DNA damage. 53BP1 requires its oligomerization domain for recruitment to sites of DNA damage and shows hallmarks of dynamic self-assembly by phase separation [25][26][27] . A reduced concentration of H4K20me2, as present in replicated nascent chromatin, therefore likely increases the threshold for efficient 53BP1 accumulation. Conversely, the high concentration of H4K20me2 in unreplicated chromatin (i.e. in G1/G0 cells and in unreplicated regions of the genome during S-phase progression), together with DNA damage-induced chromatin modifications that promote multivalent 53BP1 chromatin binding, provides a scaffold for efficient 53BP1 assembly around DNA break sites in the absence of a replicated template DNA required for HR repair. We therefore suggest that the effect of H4K20me2 dilution on 53BP1 assembly in nascent replicated chromatin functionally cooperates with the effect of H4K20me0-mediated BRCA1-BARD1 recruitment and the ensuing 53BP1 displacement, and that together they represent a dual switch to ensure that DSBs in unreplicated areas of the genome are protected from excessive DNA end resection and illegitimate recombination 28 and channeled towards NHEJ, while DSBs in replicated areas of the genome are released from the DNA end protection functions of 53BP1 and channeled towards resection and HR (Figure 7). Accordingly, the mutual antagonism between 53BP1 and BRCA1-BARD1 may be seen as a bistable system, the robustness of which is achieved by cooperative effects resulting from high affinity 53BP1 and low affinity BRCA1-BARD1 binding to unreplicated chromatin versus low affinity 53BP1 and high affinity BRCA1-BARD1 binding in replicated chromatin. This view is consistent with supraphysiological 53BP1 accumulation at damaged replicated chromatin 6 in BRCA1-deficient cells 29 , yet it suggests that even in absence of BRCA1-BARD1 the accumulation of 53BP1 is curtailed by reduced H4K20me2 in replicated chromatin. The DDR thus makes use of replication-coupled dilution of an abundant histone mark, which cells only restore in replicated chromatin when genome duplication has been completed, and enforced premature restoration of H4K20me2 during S-phase progression indeed shifts the balance towards a 53BP1-governed response 11 . In light of the growing interest to target the balance between HR and NHEJ in cancer therapy and to modify the underlying mechanisms for improved genome editing 30 , we envision the dual switch mechanism from an H4K20me2-53BP1-dominated response in unreplicated chromatin to an H4K20me0-BRCA1-BARD1-dominated response in replicated nascent chromatin to be of relevance.
More generally, we propose that, depending on the biological question, appropriate normalization becomes inevitable to interpret high-content single cell data, and that imagebased normalization to cell size, nuclear volume, DNA content, or other suitable cell cycle markers can provide additional layers of information, which may be critical for quantifying and interpreting cellular responses to stress, including genotoxic stress by irradiation, chemotherapy, or newly emerging anti-cancer drugs.   The quantification from (b) was normalized at the single cell level to DNA content to control for increasing damage load with increasing DNA amount (see Figure 2). The parameter resulting from this normalization has arbitrary units and was multiplied by a multiple of 10 to yield data that could be plotted on linear scale in the depicted range. (e) For the same cells shown in (b), the accumulated intensity of 53BP1-mScarlet foci per nucleus was normalized to the DNA content of the same nucleus.

Cell culture and treatments
Human U-2 OS cells and U-2 OS cells expressing 53BP1-mScarlet from the endogenous promoter 25 were grown under standard cell culture conditions (humidified atmosphere, 5% CO2) in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (GIBCO) and penicillin-streptomycin antibiotics. SUM149PT cells (kindly provided by Alessandro Sartori) were grown in Ham's F-12 medium (Thermo Fisher Scientific) containing 10% fetal bovine serum and 1% of penicillin-streptomycin antibiotics. All cells were maintained in a sterile cell culture environment and routinely tested for mycoplasma contamination. Irradiation was performed with a Faxitron Cabinet X-ray System Model RX-650. Transfections with Ambion Silencer Select siRNAs were performed for 72h using Lipofectamine RNAiMAX (Thermo Fisher Scientific). The following Silencer Select siRNAs were used at a final siRNA concentration of 25nM: siBRCA1 (s459) and siBARD1 (s1887).
Negative Silencer Select control Neg1 from Ambion was used as non-targeting control. For pulsed EdU (5-ethynyl-2'-desoxyuridine) (Thermo Fisher Scientific) incorporation, cells were incubated for 20 min in medium containing 10μM EdU. The Click-iT EdU Alexa Fluor Imaging Kit (Thermo Fisher Scientific) was used for EdU detection.

Quantitative image-based cytometry (QIBC)
Automated multichannel wide-field microscopy for quantitative image-based cytometry For normalization according to DNA content, measurement parameters (e.g. 53BP1 foci numbers) were divided at the single cell level by the DNA content (measured as total DAPI intensity per nucleus). The parameter resulting from this normalization has arbitrary units and was multiplied by a multiple of 10 to yield data that could be plotted on linear scale in the depicted range. Scatter plots of asynchronous cell populations were generated with Spotfire data visualization software (TIBCO). Within one experiment, similar cell numbers were compared for the different conditions. Representative scatter plots and quantifications of independent experiments, typically containing several thousand cells each, are shown.