Systematic analysis of the molecular and biophysical properties of key DNA damage response factors

  1. Joshua R Heyza
  2. Mariia Mikhova
  3. Aastha Bahl
  4. David G Broadbent
  5. Jens C Schmidt  Is a corresponding author
  1. Institute for Quantitative Health Science and Engineering, Michigan State University, United States
  2. Department of Biochemistry and Molecular Biology, Michigan State University, United States
  3. College of Osteopathic Medicine, Michigan State University, United States
  4. Department of Physiology, Michigan State University, United States
  5. Department of Obstetrics, Gynecology, and Reproductive Biology, Michigan State University, United States

Abstract

Repair of DNA double strand breaks (DSBs) is integral to preserving genomic integrity. Therefore, defining the mechanisms underlying DSB repair will enhance our understanding of how defects in these pathways contribute to human disease and could lead to the discovery of new approaches for therapeutic intervention. Here, we established a panel of HaloTagged DNA damage response factors in U2OS cells which enables concentration-dependent protein labeling by fluorescent HaloTag ligands. Genomic insertion of HaloTag at the endogenous loci of these repair factors preserves expression levels and proteins retain proper subcellular localization, foci-forming ability, and functionally support DSB repair. We systematically analyzed total cellular protein abundance, measured recruitment kinetics to laser-induced DNA damage sites, and defined the diffusion dynamics and chromatin binding characteristics by live-cell single-molecule imaging. Our work demonstrates that the Shieldin complex, a critical factor in end-joining, does not exist in a preassembled state and that relative accumulation of these factors at DSBs occurs with different kinetics. Additionally, live-cell single-molecule imaging revealed the constitutive interaction between MDC1 and chromatin mediated by its PST repeat domain. Altogether, our studies demonstrate the utility of single-molecule imaging to provide mechanistic insights into DNA repair, which will serve as a powerful resource for characterizing the biophysical properties of DNA repair factors in living cells.

Editor's evaluation

This manuscript reports valuable tools and data to study DNA repair and its regulation in life cells by generating and validating cell lines with Halo-tag fusions to the chromosomal genes encoding ATM, NBS1, MDC1, RNF168, RNF169, 53BP1, RIF1, SHLD3, REV7, SHLD2, SHLD1, and DNA-PKcs. The data establish the utility of most of the tools but remain incomplete. Conclusions from the kinetic analysis would benefit from more validation by genetic experiments and the single particle tracking analysis offers more potential for analysis.

https://doi.org/10.7554/eLife.87086.sa0

Introduction

Genomic DNA is constantly exposed to a variety of endogenous and exogenous agents that induce DNA double strand breaks (DSBs). These agents include reactive oxygen species, metabolic byproducts, and environmental carcinogens. If left unrepaired, DSBs can lead to the loss of genetic information, chromosome rearrangements, mutations, and telomere fusions that can result in the development of a wide variety of human diseases including cancer, immunodeficiencies, neurological syndromes, and premature aging disorders. In mammalian cells two main pathways exist for resolving DNA DSBs, non-homologous end joining (NHEJ) and homologous recombination repair (HR) (Scully et al., 2019). In addition, cells have evolved a complex signaling network that senses DNA lesions, called the DNA damage response (DDR). HR requires a sister chromatid to serve as a template for high-fidelity DSB repair and is limited to S- and G2-phases of the cell cycle. In contrast, NHEJ functions independently of a repair template by ligating together broken DNA ends often leading to genomic insertions and deletions. Because of the toxicity of persistent, unrepaired DNA DSBs, NHEJ functions ubiquitously throughout G1-, S-, and G2-phases of the cell cycle. While cells balance the use of NHEJ and HR particularly in S- and G2-phase, NHEJ-dependent repair represents the main pathway resolving the bulk of DSBs arising in human cells either through rapid recruitment of the core NHEJ complex (DNA-PKcs, KU70, KU80, XLF, XRCC4, and LIG4) at unresected DNA ends or after 53BP1-Shieldin mediated end fill-in of previously resected DSBs (Setiaputra and Durocher, 2019).

The DDR is critical for regulating whether DSBs are repaired via HR or NHEJ. DDR action encompasses three specific steps: DNA break detection, DDR signal amplification, and recruitment of processing enzymes and repair effectors. DSB detection can be carried out both by the Ku70/Ku80 heterodimer to directly recruit DNA-PKcs as well as ATM kinase and the MRN (Mre11, Rad50, and NBS1) complex which associate with DNA breaks leading to phosphorylation of H2AX at S139 (γH2AX) by ATM (Ciccia and Elledge, 2010). Next, DDR signal amplification is mediated by MDC1, RNF8, and RNF168 (Doil et al., 2009; Kolas et al., 2007; Lukas et al., 2004). MDC1 directly binds γH2AX through its C-terminal BRCT domains leading to recruitment of RNF8 and K63-linked polyubiquitylation of linker histone H1 (Mailand et al., 2007; Thorslund et al., 2015). RNF8-mediated histone ubiquitylation is critical for subsequent downstream recruitment of DSB repair effector proteins, such as 53BP1 and BRCA1, to the chromatin regions flanking DSBs. K63-linked ubiquitination serves as a docking site for RNF168 which further amplifies the DDR signal via its E3 ubiquitin ligase activity by depositing monoubiquitin marks onto H2A at K13 and K15 (H2AK13ubK15ub) (Mattiroli et al., 2012).

The critical step in repair pathway choice is the recruitment of either BRCA1 to promote HR or 53BP1 to facilitate NHEJ. One key regulator of this choice is RNF169 which binds H2AK13ubK15ub deposited by RNF168 (Kitevski-LeBlanc et al., 2017). RNF169 is a negative regulator of 53BP1 accumulation at DSBs and promotes BRCA1 recruitment (An et al., 2018; R. Menon et al., 2019). 53BP1 binds to H2AK15ub in addition to H4K20me2 which serves as the basis for recruitment of 53BP1 to DSBs (Wilson et al., 2016). 53BP1 binding in turn triggers recruitment and assembly of the 53BP1-RIF1-Shieldin complex where 53BP1 associates with RIF1 which in turn binds to SHLD3, REV7, SHLD2 and SHLD1 (Dev et al., 2018; Gupta et al., 2018; Noordermeer et al., 2018; Figure 1A). The Shieldin complex directly binds to ssDNA mediated by the OB-fold domains of SHLD2 (Dev et al., 2018; Gao et al., 2018; Noordermeer et al., 2018). Another important regulator of DSB repair pathway choice is the extent of end resection at the DNA break. Resection is carried out by several nucleases including CtIP and the MRN complex, Exo1, and DNA2 (Zhao et al., 2020). Extensive end resection is required for HR while limited resection favors NHEJ. One important function of the Shieldin complex is to protect DNA breaks by competing with HR factors for ssDNA binding and inhibiting access to exonucleases to prevent further resection. In addition, the Shieldin complex promotes end fill-in by recruiting Polα/Primase via the CST (CTC1, STN1, and TEN1) complex through an interaction between SHLD1 and CTC1 (Mirman et al., 2022). By preventing resection and promoting end fill-in the Shieldin complex mediates the formation of DNA ends compatible with NHEJ.

Figure 1 with 1 supplement see all
Generation of a panel HaloTagged DNA damage response proteins with CRISPR-Cas9 and homology-directed repair.

(A) Model of the DDR factors HaloTagged by genome editing and their roles in DSB repair. Red star indicates HaloTagged protein. (B) Agarose gels depicting PCR products amplified from genomic DNA showing insertion of the 3xFLAG-HaloTag into the genomic loci of each tagged DDR factor using primers oriented outside of both left and right homology arms. (WT = wildtype; KI = knock in) (C) SDS-PAGE gel showing fluorescently labeled HaloTagged proteins in each cell line after labeling with JF646 HaloTag ligand.

While we have extensive knowledge of the proteins involved in DNA repair, their genetic interactions, and biochemical activities, how their dynamic recruitment to DNA breaks controls repair pathway choice and preserves genome integrity is poorly understood. Importantly, absolute protein abundance, the mechanism by which repair factors search for breaks (e.g. 3D-diffusion, chromatin sampling/scanning), the dynamics of DNA break binding, and the sequence of repair factor recruitment to DSBs are critical determinants of repair pathway choice. In this study we have generated a collection of cell lines that express 12 HaloTagged DNA repair factors from their endogenous genomic loci. The tagged proteins encompass a variety of functions including DSB detection (ATM and NBS1), DDR signal amplification (MDC1 and RNF168) and repair effector recruitment (RNF169, 53BP1, RIF1, SHLD3, REV7, SHLD2, SHLD1, and DNA-PKcs). Using this panel of endogenously edited cell lines we systematically determined the absolute protein abundance of each protein, the kinetics of recruitment to laser-induced DSBs, and defined the diffusion dynamics and search mechanism of all factors using live-cell single-molecule imaging. We demonstrate that live cell single-molecule imaging is a highly sensitive method to detect the chromatin recruitment of DNA repair factors in response to DSB induction that is not limited to their accumulation in DNA repair foci. Furthermore, live-cell single-molecule imaging of the lowly abundant Shieldin complex components (SHLD1, SHLD2, SHLD3) demonstrates that the Shieldin complex does not exist as a preassembled complex but rather assembles at DNA lesions. Finally, live-cell single-molecule imaging reveals that MDC1 exists in a constitutive chromatin-associated state, which is mediated by MDC1’s large unstructured PST repeat region and is independent of its BRCT domain. Altogether, our work provides new insight into the molecular mechanisms of DNA repair in human cells, establishes a new approach to analyze DNA repair factor recruitment to DNA lesions in living cells, and our panel of cell lines expressing HaloTagged DNA repair factors from their endogenous loci will be a powerful resource for the DNA repair field.

Results

Generation of a panel of endogenously HaloTagged DDR proteins by genome editing

To investigate the dynamics of DNA repair proteins at the single-molecule level in living cells, we used CRISPR-Cas9 and homology-directed repair to insert a 3x-FLAG-HaloTag at their endogenous genomic loci in U2OS cells (Figure 1—figure supplement 1A). U2OS cells have been widely used as a model cell line for DNA repair studies, however, it is important to note that U2OS cells do not express ATRX, a known contributing factor to HR (Elbakry et al., 2021; Juhász et al., 2018). We selected a variety of DNA repair factors that encompass various functional steps of DNA repair and the DNA damage response (DDR) including DNA double strand break (DSB) detection (ATM & NBS1), DDR signal amplification (MDC1 & RNF168), inhibition of DNA end resection (53BP1, RNF169, RIF1, SHLD3, REV7, SHLD2, and SHLD1), as well as DSB resolution (DNA-PKcs) (Figure 1A). Proteins were tagged at either the N-Terminus (MDC1, 53BP1, SHLD3, SHLD2, SDHL1, and DNA-PKcs) or the C-Terminus (NBS1, ATM, RNF168, RNF169, RIF1 & REV7) (Figure 1—figure supplement 1B).

We generated clonal cell lines for all targeted proteins and confirmed genome editing using genomic PCR and Sanger sequencing. All cell lines were homozygously edited except for Halo-SHLD2 which had one tagged allele and a frameshift mutation in the second allele (Figure 1B and Figure 1—figure supplement 1C and E). We obtained at least two clones for all knock-ins except for NBS1, 53BP1 and SHLD2 for which we obtained a single knock-in clone expressing only the HaloTagged protein. The presence of HaloTagged protein was validated by detection of fluorescently labeled proteins in an SDS-PAGE gel after labeling with the cell permeable HaloTag ligand JF646 (Figure 1C, Figure 1—figure supplement 1D). In addition, we confirmed that all cell lines exclusively expressed the tagged protein by western blotting when antibodies for these proteins were commercially available, which is critical to assess whether the HaloTagged proteins are fully functional (Figure 1—figure supplement 1F). HaloTagged proteins validated by Western blotting were expressed at or near endogenous protein levels (Figure 1—figure supplement 1F). Despite the commercial availability of antibodies to SHLD2, SHLD1, and RNF169, we were unsuccessful at detecting these proteins at endogenous expression levels. As an alternate approach for proteins where commercial antibodies were not available, we confirmed expression of these proteins by western blotting using an antibody directed to the 3 X FLAG epitope (RNF169, SHLD3, SHLD2, and SHLD1; Figure 1—figure supplement 1D). In total, we generated a panel of 12 clonal cell lines that exclusively express HaloTagged DDR factors from their endogenous loci.

Functional validation of HaloTagged DDR proteins

To ensure that the HaloTagged DDR factors were functional, we assessed protein localization, recruitment to DNA damage induced foci, and clonogenic survival assays after treatment with the DSB-inducing drug Zeocin. To evaluate the subcellular localization of HaloTagged proteins, we imaged JF646-labeled HaloTagged proteins in live cells. As previously described, NBS1, MDC1, 53BP1, RIF1, RNF168, RNF169, and DNA-PKcs localized to the nucleus, whereas SHLD3, SHLD2, SHLD1, and REV7 were found in the cytoplasm and the nucleus (Figure 2A), suggesting that the HaloTag may not impact the proper cellular localization of these proteins (Noordermeer et al., 2018; Wilson et al., 2016). Most of these proteins also appeared to be excluded from nucleoli (Figure 2A). HaloTagging ATM at the N-terminus may lead to nuclear exclusion of the protein. To confirm that HaloTagging does not interfere with the recruitment of the tagged proteins to DNA damage sites, we analyzed their sub-cellular distribution by live-cell imaging one hour after inducing DNA DSBs with Zeocin, a radiomimetic drug chemically similar to bleomycin that also induces DNA single-strand breaks (Povirk, 1996; Povirk et al., 1977). While very few foci were detected in the absence of DNA damage in any of the cell lines, we observed dramatic increases in DNA damage induced foci formation for all proteins except for ATM and DNA-PKcs which would not be expected due to the limited number of these proteins that localize to DSBs (Figure 2A). This data confirms HaloTagged DDR proteins are capable of being recruited to DNA damage sites induced by Zeocin.

Figure 2 with 1 supplement see all
HaloTagged DDR proteins retain proper subcellular localization, foci-forming ability, and are competent for DNA repair.

(A) Representative images of JF646-labeled HaloTagged proteins in the absence or presence of Zeocin in living cells. Data presented show protein cellular localization and foci-forming ability. Scale bar = 10 μm. Images are scaled differently between untreated and treated samples to demonstrate both localization and foci-forming ability. (B) Clonogenic survival assays representing the Zeocin-sensitivity of each HaloTagged DDR cell line relative to untagged parental U2OS cells. Data presented are the results of at least three independent experiments each plated in triplicate ± S.D. Data were compared by one-way ANOVA with Dunnett’s posthoc test. * p<0.05; *** p<0.001.

Finally, we tested sensitivity to Zeocin-induced DNA damage by clonogenic survival assays of all HaloTagged DDR clones compared to untagged parental U2OS cells. Considering the size of the HaloTag (34 kDa), it is possible that it could affect the biochemical activity of its fusion partner. If the HaloTag interfered with protein function in DNA repair, we expected to observe a decrease in clonogenic survival after Zeocin treatment. For most of the HaloTagged proteins Zeocin sensitivity was indistinguishable from the parental U2OS cells, which was consistent between clones (Figure 2B and Figure 2—figure supplement 1A), confirming that these tagged proteins likely retain at least partial DNA repair function. Four cell lines were more sensitive to Zeocin than control cells including Halo-ATM, Halo-MDC1, Halo-53BP1, and Halo-SHLD2. The HaloTag-ATM cell line was as sensitive to Zeocin as U2OS cells treated with the ATM inhibitor (ATMi) (KU-55933), which could not be alleviated by moving the HaloTag to the C-terminus of ATM, demonstrating that HaloTagging ATM at either end results in a non-functional protein (Figure 2—figure supplement 1B–D). To assess whether increased sensitivity to Zeocin induced DNA damage reflected a complete loss of protein function, we established 53BP1 and MDC1 knockout cell lines using CRISPR-Cas9. The Halo-53BP1 and Halo-MDC1 cell lines had modestly increased sensitivity to Zeocin but were significantly more resistant than their knockout counterparts, demonstrating they are partially functional (Figure 2—figure supplement 1E–G). Although the Halo-SHLD2 cell line is modestly sensitive to Zeocin in clonogenic assays, it does retain proper cellular localization consistent with observations from previous studies and is successfully recruited to DNA damage sites as demonstrated by its foci-forming ability (Dev et al., 2018; Gupta et al., 2018; Noordermeer et al., 2018). Additionally, we attempted to tag Halo-SHLD2 at the C-terminus but were unsuccessful in obtaining tagged clones. It is important to consider that HaloTag-induced changes in protein function could be critical variables in interpreting experimental results. While it is unclear for what reason HaloTagging has some functional impact on MDC1, 53BP1, and SHLD2, it appears to be independent of their ability to be recruited to DSB sites, as demonstrated by each protein’s robust capacity to form foci in response to Zeocin-induced DNA damage. In summary, with the exception of ATM all of the HaloTagged proteins display expected subcellular localization within cells and are robustly recruited to DNA damage induced foci, and clonogenic assays suggest that most possess at least partial DNA repair functionality compared to their untagged counterparts.

Quantification of absolute cellular protein abundance of DNA repair factors

A key determinant of the kinetics of DDR protein recruitment to DNA damage sites is the absolute cellular concentration of the respective DNA repair factor. For example, the core NHEJ factors, DNA-PKcs, Ku70, and Ku80, are highly abundant proteins which is thought to influence their rapid recruitment to DSBs (Carter et al., 1990; Cho et al., 2022; Mimori et al., 1986). Additionally, very little is known about the absolute abundance of members of the Shieldin complex in cells which are thought to be some of the least abundant proteins in the human proteome (Gupta et al., 2018). Thus, determining the absolute cellular protein abundances is critical to establish a mechanistic, quantitative model of the DNA damage response. To measure the protein abundance of the tagged DNA repair factors, we used in-gel fluorescence and flow cytometry-based methodologies. For in-gel fluorescence quantification of the HaloTagged proteins, we modified a method originally described by Cattoglio et al., 2019. To perform these experiments, we generated a standard curve using known quantities of recombinant 3XFLAG-HaloTag labeled with JF646 and cell lysates from a specific number of U2OS cells (Figure 3A). Two independent clones of each cell line were labeled with 500 nM JF646, which was ~10 x higher than the saturating concentration for the most abundant protein, REV7 (Figure 3—figure supplement 1A). Because each protein migrates differently on SDS-PAGE gels based on its size and amino acid composition, we considered the possibility that this may influence the fluorescence of each sample relative to the purified 3XFLAG-HaloTag. To account for differences in the fluorescence signal cause by different migration rates and patterns we cleaved the HaloTag from each protein prior to SDS-PAGE using TEV protease to create an adjustment factor for each protein (Figure 3—figure supplement 1B). We obtained a TEV correction factor for each protein except RNF168, RNF169, and MDC1 which all appeared to be degraded after cell lysis which could not be alleviated by including a protease inhibitor cocktail (Figure 3—figure supplement 1C). Correction factors ranged 1.04 for SHLD1 to 3.04 for SHLD2 meaning that cleaved HaloTag fluorescence intensity was 1.04x – 3.04x greater for the cleaved HaloTag, than the full-length fusion protein (Figure 3—figure supplement 1D).

Figure 3 with 1 supplement see all
HaloTag enables quantification of absolute cellular protein abundances.

(A) Example image of in-gel fluorescence of JF646-labeled HaloTagged proteins. (B) Comparison of JF646 fluorescence intensity values (normalized to NBS1-Halo) between in-gel fluorescence after applying the TEV correction factor and flow cytometry. White columns indicate unadjusted samples because of the inability to accurately determine a TEV correction factor due to protein degradation. Data are presented as the mean of ≥ three independent experiments ± S.D. (C) Plot representing the correlation of normalized JF646 fluorescence intensities ± S.D. for each protein between TEV-corrected in-gel fluorescence and flow-cytometry. Data were analyzed by Pearson’s correlation coefficient. White points indicate those proteins for which a TEV-correction factor could not be accurately determined. Red line represents an interpolated standard curve.

After correcting for differences in fluorescence intensity cause by SDS-PAGE migration patterns, HaloTagged protein abundance ranged from ~1600 (ATM) – 180,000 (REV7) molecules per cell (Table 1), with ATM (~1600–4200 molecules per cell), SHLD1 (~7300–8100 molecules per cell), SHLD2 (~7300 molecules per cell), and SHLD3 (~4000–5600 molecules per cell) clones having the lowest expression level (Table 1). Considering a large proportion of SHLD1, SHLD2, and SHLD3 protein is localized to the cytoplasm, the nuclear protein abundance for each of these proteins is even lower which may influence the kinetics of Shieldin complex recruitment in 53BP1-dependent NHEJ. We observed higher protein abundances for factors involved in initial break detection including NBS1 (~16,000 molecules per cell) and DNA-PKcs (~86,000–91,000 molecules per cell) and those contributing to DDR signal amplification including MDC1 (~2400–13,000 molecules per cell), RNF168 (~22,000–28,000 molecules per cell), and RNF169 (~18,000–19,000 molecules per cell) (Table 1). The highest protein abundances were observed for 53BP1 (~40,000 molecules per cell), RIF1 (~58,000–64,000 molecules per cell), REV7 (~78,000–180,000 molecules per cell), and DNA-PKcs (~86,000–91,000 molecules per cell) (Table 1). Importantly, the differences in absolute protein number between independent genome edited clones could be the consequence of either a different number of alleles being modified with the HaloTag, considering the possibility of inherent heterogeneity in karyotype within U2OS cells, or alternatively could be due to variations in protein expression within different clones. As a complementary approach, we used flow cytometry to measure the fluorescence intensity of JF646-labeled HaloTagged proteins in cells (Figure 3—figure supplement 1E). The protein abundance values we obtained by in-gel fluorescence were compared to flow-cytometric quantifications of mean fluorescence intensity for each cell line. Flow-cytometry does not provide an absolute protein number, but instead measures relative protein abundance between the cell lines expressing different HaloTagged proteins. The relative fluorescence levels detected by flow cytometry corresponded well with the relative absolute protein abundance determined by in-gel fluorescence (Figure 3B and C). Furthermore, results with this complementary approach were comparable to in-gel fluorescence for proteins where a TEV correction factor could not be accurately determined indicating that a TEV correction would have only led to modest changes in overall abundance of these proteins.

Table 1
Absolute protein abundances of HaloTagged DNA repair proteins in U2OS cells.

Left column: Absolute protein abundance of HaloTagged proteins determined by in-gel fluorescence after adjusting with the TEV correction factor calculated for each protein. An asterisk (*) indicates the three proteins for which an accurate TEV correction factor could not be generated. For TEV-corrected samples, the S.D. includes propagated error. Right column: Determination of absolute protein abundances of untagged DNA repair proteins in parental U2OS cells by applying a Western blot correction factor to adjust for differences in expression between HaloTagged and untagged protein. For western blot corrected samples, S.D. includes propagated error. An asterisk (*) denotes samples where the Western correction was applied to samples that were not TEV corrected. No Data indicates the absence of a commercially available antibody or an antibody that detects endogenous levels of protein expression.

ProteinHaloTag Protein ± S.D.Western Correction ± S.D.
NBS115,846±954718,055±11,180
ATM C14161±18974738±2379
ATM C21578±9453115±2365
MDC1 C113,352±6001*16,030±7631*
MDC1 C22434±15063390±2180*
RNF168 C127,787±7241*4585±1819*
RNF168 C221,531±4801*4678±1145
RNF169 C119,013±3759*No Data
RNF169 C218,413±3119*No Data
53 BP140,162±11,49812,109±5792
RIF1 C157,486±28,9709382±5765
RIF1 C264,358±37,50410,398±7506
SHLD3 C15555±1913No Data
SHLD3 C23953±2187No Data
REV7 C1178,131±40,75739,803±10,299
REV7 C278,340±25,16166,790±42,830
SHLD27257±3104No Data
SHLD1 C18124±4104No Data
SHLD1 C27257±3103No Data
DNA-PKcs C190,899±26,030115,143±46,933
DNA-PKcs C285,786±23,508138,704±71,488

Finally, using western blotting for each protein where an antibody was commercially available and could detect endogenous levels of protein expression, we quantified the concentration of each HaloTagged protein relative to untagged wild-type protein in parental U2OS cells. Expression of HaloTagged proteins relative to untagged protein in U2OS cells ranged from 0.62 x and 0.79 x for HaloTagged DNA-PKcs clones to 6.13 x and 6.19 x for RIF1 clones with 53BP1, RNF168, and one REV7 clone also being expressed at higher levels upon HaloTagging, potentially as a consequence of increased protein stabilization or increased protein production (Figure 3—figure supplement 1F). We used the relative expression of each protein compared to untagged protein to calculate the number of molecules per cell for each protein in parental U2OS cells (Table 1). After applying this adjustment factor, DNA-PKcs was the most abundant protein ranging from ~71,000 to 120,000 molecules per cell. Conversely, ATM and RNF168 had the lowest protein abundances ranging from ~3100 to 4700 molecules per cell. Additionally, the calculated number of molecules per cell for both clones expressing each tagged protein fell within the standard deviation of each other with the exception of MDC1. This high variability of calculated molecules per cell between MDC1 clones is likely attributable to inconsistent signals in western blots which was not alleviated by optimizing multiple wet transfer conditions or by using two separate MDC1 antibodies. In summary, in-gel fluorescence enabled quantification of absolute protein abundance of HaloTagged proteins in U2OS cells with a dynamic range capable of detecting both lowly and highly expressed proteins.

Kinetics of recruitment of HaloTag DDR proteins to sites of laser-induced microirradiation

After validating that our panel of HaloTagged DDR factors are functional and proficient in DNA repair, we monitored the kinetics of recruitment for each factor to sites of laser-induced DNA damage. While laser-induced microirradiation (LMI) induces DNA DSBs, it also causes a variety of other types of DNA damage including single-strand breaks and base damage and likely induces DNA damage at levels typically not encountered by human cells (Holton et al., 2017; Muster et al., 2017). Even considering these important drawbacks, LMI has been widely used as one tool to monitor recruitment kinetics of DNA repair proteins to spatially and temporally controlled sites of DNA damage. Most previous studies that monitored the kinetics of recruitment of DNA repair factors to laser-induced DNA damage sites have used transgene expression of fluorescently tagged proteins, which could influence their recruitment kinetics due to changes in protein levels and competition with the untagged endogenous protein. The use of HaloTagged proteins expressed from their endogenous loci allowed us to more accurately determine the relative recruitment kinetics of each DNA repair factor to laser-induced DNA lesions. HaloTagged DNA repair factors were labeled with JFX650 and cells were pre-sensitized to LMI with Hoechst (1 μg/mL). LMI reproducibly induced robust protein recruitment for each DNA repair factor (Figure 4A, Figure 4—figure supplement 1A, Videos 110). To compare the relative recruitment kinetics of the DNA repair factors we determined the time to half-maximal accumulation (t1/2) (Figure 4—figure supplement 1B). DNA-PKcs and NBS1 accumulated rapidly after LMI, with a t1/2 = 22.1 s and t1/2 = 31.3 s, respectively, consistent with their known roles in the early steps of the DNA damage response (Figure 4B and Figure 4—figure supplement 1B). MDC1 (t1/2 = 76.5 s) and RNF168 (t1/2 = 69.4 s) arrived at similar times, consistent with rapid ubiquitin deposition at DSBs dependent upon MDC1 mediated RNF8 and RNF168 recruitment (Figure 4B and Figure 4—figure supplement 1B). RNF169 (t1/2 = 185.8 s) was significantly delayed compared to RNF168 (Figure 4B and Figure 4—figure supplement 1B), indicating that RNF169 requires chromatin modification by RNF168 prior to its recruitment. 53BP1 and RIF1 both accumulated slowly at LMI sites with comparable kinetics (t1/2 = 669.0 s, t1/2 = 550.8 s, respectively) (Figure 4C and Figure 4—figure supplement 1B). REV7 (t1/2 = 90.8 s) and SHLD3 (t1/2 = ~1168 s), which are thought to form a complex reached their half-maximal accumulation at LMI induced DNA lesions with distinct kinetics, although REV7 possesses additional functions in DNA repair beyond the Shieldin complex, which could lead to the observed recruitment kinetics but was not further analyzed in this study (Figure 4C and Figure 4—figure supplement 1B). Importantly, we did not analyze recruitment kinetics of REV7-Halo in the absence of SHLD3. In addition, SHLD2 (t1/2 = 287.1 s) is recruited more rapidly than 53BP1 and RIF1, which suggests that either SHLD2 is not strictly dependent on these proteins and can be recruited to DSBs by its association with ssDNA, or that at the time SHLD2 begins to accumulate at DSB sites there is already sufficient 53BP1, RIF1, and SHLD3 for SHLD2 recruitment. To distinguish between these possibilities, we knocked out SHLD3 in Halo-SHLD2 cells and imaged SHLD2 in the presence or absence of Zeocin-induced DSBs in live-cells (Figure 4—figure supplement 1C). While SHLD3 wildtype cells supported SHLD2 foci-formation after Zeocin treatment, loss of SHLD3 completely eliminated SHLD2 foci-formation, consistent with the well-known genetic dependency of SHLD2 on SHLD3 and with the hypothesis that at the time SHLD2 accumulation begins sufficient SHLD3 is present at DNA breaks to initiate SHLD2 recruitment. To confirm this by comparing the absolute amount of Halo-SHLD2 and Halo-SHLD3 recruited to laser induced DNA damage sites, we repeated the experiments using identical imaging conditions and without intensity normalization. In these experiments Halo-SHLD2 and Halo-SHLD3 were simultaneously recruited to DNA lesions in comparable amounts (Figure 4—figure supplement 1D). After Halo-SHLD2 accumulation plateaued, Halo-SHLD3 continued to accumulate in excess of Halo-SHLD2 (Figure 4—figure supplement 1D). Finally, we also observed recruitment of SHLD1 to laser-induced DSBs but were limited to acquiring images every ten minutes due to technical issues with photo-bleaching of the fluorescence signal at faster imaging rates that resulted from low SHLD1 expression and slow recruitment to LMI sites (Figure 4—figure supplement 1E). Interestingly, SHLD1 is recruited in very low amounts and had the appearance of single particles enriched in the area of laser microirradiation, but the signal intensity was insufficient to reliably determine the recruitment kinetics of SHLD1. The recruitment kinetics of Halo-RNF168 and Halo-RNF169 are comparable to those observed in a previous study that systematically analyzed DNA repair factor recruitment to laser induced damage sites (Aleksandrov et al., 2018). The t1/2 data for Halo-MDC1 and Halo-53BP1 we observed differed by approximately 2-fold from those reported by Aleksandrov et al., 2018. These differences could be a consequence of the expression method (Bacmid transgene expression vs. genome editing) or the species from which the DNA repair factor was derived (mouse 53BP1 vs. human 53BP1). In summary, we have determined the recruitment kinetics of the HaloTagged DNA repair factors expressed at endogenous or near endogenous expression levels. The wide range of observed kinetics provides insights into the relative recruitment of the analyzed DDR factors to DSBs and the stoichiometry of Shieldin complex subunits.

Figure 4 with 1 supplement see all
Kinetics of HaloTagged DDR proteins recruitment to sites of laser microirradiation induced DNA breaks.

(A) Representative images of NBS1-Halo (JFX650) recruitment to laser-induced DSBs over time (Scale bar = 10 μm). (B) Normalized recruitment kinetics of HaloTagged DNA-PKcs, NBS1, MDC1, RNF168, and RNF169 proteins to laser-induced DSBs. (C) Normalized recruitment kinetics of HaloTagged 53BP1, RIF1, REV7, SHLD2, and SHLD3 proteins to laser-induced DSBs. Data are presented as the average increase in fluorescence post-laser microirradiation normalized to the brightest average frame for each movie. n=8–13 individual cells analyzed for each HaloTag cell line.

Video 1
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged DNA-PKcs to DNA DSBs after laser microirradiation.

Images were acquired at one frame per second. 170x170 pixels with a pixel size of 0.16 μm.

Video 2
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged NBS1 to DNA DSBs after laser microirradiation.

Images were acquired at one frame per second. 170x170 pixels with a pixel size of 0.16 μm.

Video 3
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged MDC1 to DNA DSBs after laser microirradiation.

Images were acquired at one frame per second. 170x170 pixels with a pixel size of 0.16 μm.

Video 4
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged RNF168 to DNA DSBs after laser microirradiation.

Images were acquired at one frame per second. 170x170 pixels with a pixel size of 0.16 μm.

Video 5
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged RNF169 to DNA DSBs after laser microirradiation.

Images were acquired at one frame per second. 170x170 pixels with a pixel size of 0.16 μm.

Video 6
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged 53BP1 to DNA DSBs after laser microirradiation.

Images were acquired at one frame every 10 s. 170x170 pixels with a pixel size of 0.16 μm.

Video 7
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged RIF1 to DNA DSBs after laser microirradiation.

Images were acquired at one frame every 10 s. 170x170 pixels with a pixel size of 0.16 μm.

Video 8
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged SHLD3 to DNA DSBs after laser microirradiation.

Images were acquired at one frame every 10 s. 170x170 pixels with a pixel size of 0.16 μm.

Video 9
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged REV7 to DNA DSBs after laser microirradiation.

Images were acquired at one frame every two seconds. 170x170 pixels with a pixel size of 0.16 μm.

Video 10
Representative movie demonstrating recruitment of JFX650-labeled 3xFLAG-HaloTagged SHLD2 to DNA DSBs after laser microirradiation.

Images were acquired at one frame every five seconds. 170x170 pixels with a pixel size of 0.16 μm.

Single-molecule live-cell imaging of HaloTagged DDR proteins

The diffusion dynamics of DNA repair factors in the nucleus are a key determinant of their rapid and specific recruitment to DNA lesions. Live-cell single-molecule imaging makes it possible to analyze the diffusion dynamics, chromatin binding, and recruitment to sites of DNA damage of individual DNA repair proteins in their endogenous context. To carry out single-molecule imaging of the HaloTagged DNA repair factors, we used a combination of sparse HaloTag labeling and highly inclined laminated optical sheet (HILO) microscopy (Tokunaga et al., 2008), and imaged cells at 138 frames per second to capture rapid diffusion dynamics in both unperturbed conditions and after treatment with Zeocin (Videos 1121). Single particles were automatically detected in each frame, linked into trajectories using the multi-target tracking algorithm (Sergé et al., 2008), and step size distributions were analyzed using the Spot-On tool to extract the diffusion coefficients of freely diffusing (Dfree) and chromatin bound molecules (Dbound), as well as the fraction of particles that are associated with chromatin (Fbound) (Figure 5A; Hansen et al., 2018). As model proteins for freely diffusing and chromatin bound factors, we transiently expressed and imaged the 3XFLAG-HaloTag fused to a 3 x Nuclear Localization Sequence (3XNLS) and HaloTagged histone H2B, respectively (Figure 5—figure supplement 1A, Video 22). 3XFLAG-HaloTag-3XNLS (Dfree = 3.9 μm2/s, Fbound = 16%) represents the lower bound for static particles and the upper bound for the free diffusion coefficient, while Halo-H2B (Fbound = 66%) represents the upper bound for the fraction of bound particles since histone H2B is an integral component of chromatin. The fraction of bound molecules (Fbound) values for Halo-NLS and Halo-H2B were similar to those previously reported in Hansen et al., 2018, which did not include a 3XFLAG tag fused to the HaloTag. Conversely, the free diffusion coefficients (Dfree) values we report were lower than previously reported, due to the well described impact of the higher labeling density used in our experiments on diffusion coefficient measurements of highly dynamic proteins (Hansen et al., 2018). For all proteins, we only analyzed nuclear particle trajectories and assumed a two-state diffusion model where particles either freely diffuse or are chromatin bound (Figure 5—figure supplement 1B). Importantly, determining Dfree, Dbound, and Fbound for individual cells (Figure 5A–C, Figure 5—figure supplement 1C), and combining the step size measurements from all cells for three experimental replicates (Figure 5—figure supplement 1D) lead to similar results, indicating that our analysis approach is robust. The diffusion coefficient for freely moving particles ranged from Dfree = 1.0–3.7 μm2/s, which are all lower than the diffusion coefficient measured for the HaloTag-3XNLS (Figure 5B). The diffusion coefficients measured for SHLD1 (Dfree = 3.7), SHLD2 (Dfree = 1.8), and SHLD3 (Dfree = 2.3) were significantly different from each other. This suggests that the shieldin complex is not preassembled in the nucleoplasm, which is consistent with the distinct recruitment kinetics to LMI induced sites of DNA damage we observed for the shieldin complex components (Figure 4C). Furthermore, intensity profiles of MDC1 and RIF1 trajectories provided confidence that the vast majority of particle localizations represent single fluorophores and not multiple labeled molecules co-diffusing or intersecting which would have additive effects on intensity values (Figure 5—figure supplement 1E).

Video 11
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged NBS1 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 12
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged MDC1 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 13
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged RNF168 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 14
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged RNF169 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 15
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged 53BP1 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 16
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged RIF1 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 17
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged REV7 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 18
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged SHLD3 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 19
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged SHLD2 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 20
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged SHLD1 labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Video 21
Representative live-cell single-molecule imaging movies of untreated and zeocin-treated U2OS cells expressing 3xFLAG-HaloTagged DNA-PKcs labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Figure 5 with 1 supplement see all
HaloTagged DDR proteins exhibit distinct nuclear diffusion and chromatin binding characteristics.

(A) Graphical representation of the workflow used for live-cell single-molecule imaging of HaloTagged DDR proteins. (B) Diffusion coefficients for freely diffusing HaloTag DDR proteins present in at least three consecutive frames in untreated conditions and post-Zeocin exposure. Values plotted indicate the Dfree for all analyzed tracks per cell with each dot indicating a separate cell that was analyzed. Live-cell single-molecule imaging was performed over 3–4 separate days imaging at least 20 individual cells per condition per experimental replicate (n≥60 cells total for each protein and condition representing three to four independent experiments). Red bar = median. (C) Diffusion coefficients of the bound fraction of HaloTag DDR proteins present in at least three consecutive frames in untreated conditions and post-Zeocin exposure. Values plotted indicate the Dbound for all analyzed tracks per cell with each dot indicating a separate cell that was analyzed. Live-cell single-molecule imaging was performed over 3–4 separate days imaging at least 20 individual cells per condition per experimental replicate (n≥60 cells total for each protein and condition). Red bar = median. (D) Plot of the Fraction Bound for each HaloTag DDR protein under each condition that were analyzed using a two-state model of diffusion. Each dot represents the fraction bound of each protein in an individual cell (n≥60 cells for each protein and condition). Red bar = median. Differently shaded points indicate data collected from separate biological experiments. Data were analyzed by two-way ANOVA with Tukey’s posthoc test. n.s.=not significant. ** p=0.004; **** p<0.0001.

Video 22
Representative live-cell single-molecule imaging movies of U2OS cells transiently expressing HaloTag-H2B and HaloTag-3xNLS labeled with JFX650 and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

The diffusion coefficient for bound particles in unperturbed cells ranged from Dbound = 0.02–0.10 μm2/s, which are all faster than Halo-H2B (Figure 5C). One would predict to observe differences in the Dbound between factors that are physically incorporated within nucleosomes (e.g. H2B), bind modified histone tails (e.g. MDC1), or accumulate into phase separated compartments (e.g. 53BP1). Consistent with this expectation, Halo-H2B had a Dbound of 0.012 μm2/s, while the diffusion coefficient for bound MDC1 molecules was two-fold higher (Dbound = 0.024 μm2/s). On the other hand, 53BP1 which can diffuse within phase-separated compartments had the fastest diffusing bound particles (Dbound = 0.099 μm2/s). These data point to the power of this method to detect even small differences in the behavior of DNA repair factors when bound to chromatin. We observed a wide range of the fraction of molecules bound to chromatin (Fbound = 23–69%) in unperturbed cells (Figure 5D). Importantly, the chromatin bound fraction of all DNA repair factors analyzed was significantly higher than that of HaloTag-3XNLS, which likely reflects their intrinsic propensity to associate with chromatin (Figure 5D). Alternatively, the chromatin bound molecules could be the result of low levels of DNA lesions present in control cells. Strikingly, the chromatin bound fraction of MDC1 (Fbound = 69%) and RIF1 (Fbound = 68%) in unperturbed cells was comparable to that of HaloTag-H2B (Fbound = 66%), suggesting that these DNA repair factors are constitutively associated with chromatin and may search for DNA breaks by local chromatin scanning (Figure 5D).

After Zeocin treatment, we observed a significant increase in the fraction of static particles for RNF169, 53BP1, SHLD2, SHLD3, and DNA-PKcs (Figure 5D), consistent with their recruitment to DNA breaks. Since MDC1 and RIF1 were largely chromatin associated in control cells it is unsurprising that we did not observe a further increase in their overall chromatin association after zeocin treatment (Figure 5D). It is important to note that a change in the fraction bound requires that a significant percentage of the respective DNA repair factor has to be recruited to DNA lesions, which depends on the abundance of the repair factor relative to the number of DSBs induced. NBS1, REV7, and RNF168 are among the most abundant proteins analyzed, which could explain why their Fbound did not significantly increase after zeocin treatment (Figure 5D). Additionally, HaloTagging RNF168 leads to expression ~4–6 x that of wild-type protein in U2OS cells which could also contribute to our inability to detect significant changes in protein binding after Zeocin treatment. Despite its low abundance SHLD1 we were not able to detect recruitment of SHLD1 to chromatin after DSB induction using this approach (Figure 5C), consistent with the very limited and slow recruitment we observed in the LMI experiments described above (Figure 4—figure supplement 1D). Importantly, for all factors that showed an increase in the Fbound after zeocin treatment, we observed a decrease in the Dfree (Figure 5B and D), which is likely the result of a systematic error as a consequence of the global step size fitting in Spot-On.

While these observations report on global changes in DDR protein behavior, they did not distinguish between how particles behave within or outside of repair foci, which have been traditionally used as a marker of DNA repair factor recruitment to DNA lesions. By combining sparse labeling of Halo-53BP1 by JFX650 (pseudo-colored green) with subsequent quantitative labeling using the spectrally distinct JF503 (pseudo-colored magenta) we could simultaneously image single 53BP1 molecules and DNA repair foci marked by 53BP1 accumulation at Zeocin-induced DSBs (Figure 5—figure supplement 1F; Videos 23 and 24). We observed highly static and dynamic 53BP1 particles within 53BP1 foci consistent with the relatively high Dbound we observed for 53BP1 (Figure 5—figure supplement 1G, Videos 2324). Static 53BP1 molecules within DNA repair foci likely represent molecules directly bound to chromatin while dynamic 53BP1 molecules are likely recruited by the formation of phase-separated 53BP1 compartments (Figure 5—figure supplement 1G, Videos 2324), as observed by others (Kilic et al., 2019). We also observed static 53BP1 particles not associated with DNA repair foci which could represent the initial recruitment of 53BP1 to chromatin prior to the formation of a phase-separated compartment or DNA lesions that are rapidly repaired and do not mature into a detectable DNA repair focus (Figure 5—figure supplement 1G, Videos 2324). In addition, we detected the transition of freely diffusing 53BP1 molecules in and out of DNA repair foci (Figure 5—figure supplement 1G, Videos 2324). To further dissect the dynamics of Halo-53BP1 within and outside of DNA repair foci, we used a mask of DNA repair foci to split Halo-53BP1 trajectories into tracks that overlapped with DNA repair foci and those that did not (Figure 5—figure supplement 1F). The trajectories were then analyzed with Spot-On using a three-state model assuming Halo-53BP1 molecules could either be freely diffusing, directly bound to chromatin, or part of a phase-separated DNA repair focus, which would lead to an intermediate diffusion coefficient. Within DNA repair foci 29% of 53BP1 molecules moved with a slow diffusion coefficient (Dfree2=0.6 µm2/s) compared to 17% of 53BP1 trajectories that did not overlap with DNA repair foci (Figure 5—figure supplement 1H). This observation is consistent with the presence of two types of 53BP1 molecules within DNA repair foci, one that is directly associated with chromatin and a second that is potentially recruited via a phase-separation mechanism. Taken together, combined imaging of single-particles and DNA repair foci could be a powerful tool enabling analyses to specifically monitor or differentiate between protein behaviors inside and outside of repair foci.

Video 23
Representative live-cell single-molecule imaging movie of Zeocin-treated U2OS cells expressing endogenous 3xFLAG-HaloTag-53BP1 sparsely labeled with JFX650 (pseudo-colored green) acquired at 202 frames per second and overlayed with a Z-projected image of 53BP1 foci which were densely labeling with JF503 (pseudo-colored magenta).

Pixel size = 0.108 μm.

Video 24
A second representative live-cell single-molecule imaging movie of Zeocin-treated U2OS cells expressing endogenous 3xFLAG-HaloTag-53BP1 sparsely labeled with JFX650 (pseudo-colored green) acquired at 202 frames per second and overlayed with a Z-projected image of 53BP1 foci which were densely labeling with JF503 (pseudo-colored magenta).

Pixel size = 0.108 μm.

In summary, live-cell single-molecule imaging provides insight into the molecular mechanism by which DNA repair factors search for DNA lesions, reports on complex formation, and allows the analysis of their recruitment to DNA lesions. In addition, our observations demonstrate that DNA repair factor recruitment to chromatin is not strictly confined to DNA repair foci and therefore could report on DNA repair events that were previously not detectable.

Residence time analysis of chromatin-bound DNA repair factors

To gain further insight into the biochemical properties of the DNA repair factors bound to chromatin, we analyzed the residence time of static single-particle trajectories. Accurate residence time analysis requires continuous tracking of bound molecules without any gaps in the trajectories. To robustly track long-lasting binding events, we averaged the intensity values of 10 consecutive imaging frames, which amplifies the signal of static molecules and reduces the intensity of mobile particles because their signal is spread out over multiple pixels over the course of 10 time points (Figure 6A). Single-particles were then tracked by constraining the diffusion coefficient to only detect chromatin bound molecules (Video 25). Importantly, when analyzing long-lived single-molecule fluorescence signals it is critical determine the contribution of photo-bleaching to signal disappearance. To assess the photo-bleaching rate under our imaging conditions, we analyzed the integrated the nuclear signal intensity of cells expressing Halo-H2B over time, which limits the contribution of protein dissociation and diffusion of fluorescent molecules out of the focal plane to the loss of fluorescence signal over time. The fluorescence half-life of Halo-H2B was 7.7 seconds under our imaging conditions (Figure 6B), which sets the upper limit for our residence time analysis. We used an aggregate data set including residence times from three biological replicates (>60 cells per cell line), since individual replicates lead to highly reproducible results (Figure 6—figure supplement 1A). The residence time for all factors analyzed, including Halo-H2B, fit well to two exponential decay functions (Figure 6B, Figure 6—figure supplement 1B), indicating all factors exhibit transient and stable binding to chromatin. Importantly, the half-life of stable single Halo-H2B molecules binding to chromatin matched the half-life observed of total nuclear fluorescence, indicating that our measurement of the photo-bleaching rate is accurate and H2B residence time exceeds the 22 s time frame of our experiment (Figure 6B). All DNA repair factors analyzed dissociated from chromatin more rapidly than histone H2B with half-lives ranging from t1/2 = 2.4 seconds for Halo-53BP1 to t1/2 = 5.8 s for Halo-MDC1 (Figure 6C). The induction of DNA damage using zeocin increased to residence time of Halo-53BP (t1/2 ZEO = 3.6 s), Halo-SHLD2 (t1/2 ZEO = 4.9 s), and Halo-SHLD3 (t1/2 ZEO = 5.8 s) (Figure 6C), consistent with the increase in the chromatin bound fraction of these proteins observed in our Spot-On analysis (Figure 5D). In contrast, the residence time of the remaining factors was unchanged in the presence of zeocin, which suggests that the biochemical interactions that trigger their chromatin association are comparable under both experimental conditions. Importantly for all factors analyzed a fraction of molecules were present for the entire imaging time (MDC1 F>22s=3%, 53BP1 F>22s=0.2%, Figure 6B). We failed to assess the residence time of Halo-DNA-PKcs because the high concentration of rapidly diffusing Halo-DNA-PKcs molecules lead to high background signal, making it impossible to reliably track statically bound Halo-DNA-PKcs molecules. In total, these results demonstrate that DNA break induction by zeocin and the associated changes in chromatin state (histone modifications, DNA break resection) increase the affinity of the association of 53BP1, SHLD2, and SHLD3 with DNA breaks. It is important to note that due to photo-bleaching and potential drift of the bound molecules out of the focal plane of the objective these experiments underestimate the absolute time of chromatin binding of the proteins analyzed.

Figure 6 with 1 supplement see all
Residence time analysis of DNA repair factors in response to DNA damage.

(A) Kymographs of single-molecule imaging movies for HaloTagged DNA repair factors. Movies were acquired with 7.2ms exposure times and to amplify long lasting interactions the intensity of 10 consecutive frames was averaged. (B) Residence time (track length) distribution of long-lasting chromatin binding events of DNA repair factors displayed as survival probability after the elapsed time. (C) Residence time (track length) distribution of long-lasting chromatin binding for the indicated DNA repair factors displayed as survival probability after the elapsed time. Aggregated data from 3 to 4 separate days imaging at least 20 individual cells per condition per experimental replicate (n≥60 cells total for each protein and condition) was fit with two exponential decay functions (R2 >0.99 for all proteins) and half-lives and fractions of slow and fast decaying components are reported.

Video 25
Representative movie demonstrating robust tracking of bound particles after averaging the image intensity over 10 frames and constraining the expected diffusion coefficient of bound particles to Dmax = 0.05 µm2/s.

Images acquired at 138 frames per second and the frame rate was reduced to 13.8 frames per second by the averaging over 10 consecutive frames. Pixel size = 0.16 μm.

Live-cell single-molecule imaging reveals MDC1’s constitutive chromatin interaction is mediated by its PST repeat domain

Our live-cell single-molecule imaging revealed that MDC1 is constitutively chromatin associated. MDC1 contains two domains that have been implicated in chromatin binding: The BRCT domains that bind to γH2AX (Stucki et al., 2005), and its 13 PST repeats, which have been proposed to associate with the nucleosome acidic patch (Salguero et al., 2019). To dissect the relative contribution of the BRCT and PST repeat domains to the constitutive chromatin association of MDC1, we transiently expressed HaloTagged Halo-MDC1 wildtype (WT), PST deletion (ΔPST), or BRCT deletion (ΔBRCT) mutants in MDC1 knockout (ΔMDC1) cells (Figure 7A–B). We first assessed the ability of these MDC1 variants to localize to DNA damage induced foci. WT MDC1, and MDC1 ΔPST formed foci after zeocin treatment, while MDC1 ΔBRCT was not recruited to DNA damage induced foci (Figure 7C), consistent with previous observations that BRCT-γH2AX binding is essential for MDC1 foci formation and does not require the PST domain (Stucki et al., 2005). This suggests that the BRCT domains of MDC1 are the primary driver of MDC1 recruitment to DNA lesions. Next, we performed live-cell single-molecule imaging to define the contribution of the BRCT and PST repeat domains to the constitutive chromatin association we observed in unperturbed cells (Figure 7D, Figure 7—figure supplement 1A–B, Video 26). WT MDC1 transiently expressed in ΔMDC1 behaved identically to endogenously tagged MDC1 (Figure 7D, Figure 7—figure supplement 1A–B, Figure 5B–C), and deletion of the BRCT domains did not alter the diffusion behavior of MDC1 (Figure 7D, Figure 7—figure supplement 1A–B). In contrast, deletion of the PST repeat domain almost completely eliminated the fraction of MDC1 molecules associated with chromatin (Figure 7D, Figure 7—figure supplement 1A–B). This suggests that the PST repeat domain of MDC1 mediates the constitutive chromatin association of MDC1 even in the absence of DNA damage. Because a high fraction of MDC1 particles are chromatin-bound in unperturbed cells, we suspected that this interaction masks the ability to detect γH2AX-bound MDC1 after Zeocin exposure in single-molecule imaging experiments (Figure 5D). Therefore, we analyzed the single-molecule dynamics of the MDC1 deletion mutants in the presence or absence of Zeocin (Figure 7—figure supplement 1C). Consistent with our earlier observations, no changes in the bound fraction of MDC1 were detected after Zeocin treatment in cells expressing full-length MDC1 or the BRCT deletion mutant (Figure 7—figure supplement 1C). In contrast, upon deletion of the PST repeat region, we observed a significant increase in chromatin-bound MDC1 after Zeocin exposure (Figure 7—figure supplement 1C). This observation supports the conclusion that the PST-mediated constitutive interaction of MDC1 with chromatin masked the ability to detect BRCT-domain mediated binding of MDC1 to γH2AX at DNA breaks by live-cell single-molecule imaging. To analyze the contribution of the PST- and BRCT-domains to DSB repair, we used CRISPR-Cas9 to knock-in the DR-GFP HR reporter into the AAVS1 locus in MDC1 knockout cells which we confirmed by genomic DNA PCR (Figure 7—figure supplement 1D; Pierce et al., 1999). The DR-GFP HR reporter uses a double stranded break induced in a non-functional GFP gene by the restriction enzyme I-SceI to assess repair by HR through reconstituting a functional GFP gene by a repair donor contained in the reporter cassette that lacks a functional promotor (Pierce et al., 1999). We transiently transfected HaloTagged MDC1 variants and the I-SceI expression plasmid into the DR-GFP reporter cells in which endogenous MDC1 was knocked out and used flow cytometry to detect GFP expressing cells which resulted from HR mediated repair of the GFP reporter. Expression of HaloTagged WT MDC1 led to a significantly higher number of GFP positive cells compared to MDC1 knockout cells, consistent with HaloTagged MDC1 being capable of supporting HR (Figure 7—figure supplement 1E). In contrast, HR efficiency was significantly reduced to a similar degree when MDC1 lacking the PST- or BRCT-domains was expressed, which was previously documented by others (Xie et al., 2007). Additionally, we used immunofluorescence to determine the ability of HaloTagged MDC1 variants to support recruitment of 53BP1 to Zeocin-induced DSBs. Compared to untreated cells, WT and ΔPST MDC1 were robustly recruited into DNA repair foci, while inhibiting γH2AX binding by deleting the BRCT domain prevented MDC1 foci formation consistent with previous observations (Figure 7—figure supplement 1F; Xie et al., 2007). Additionally, HaloTagged WT and ΔPST MDC1 were sufficient to promote recruitment of 53BP1 to DSB sites, while 53BP1 recruitment was defective in ΔMDC1 cells or cells expressing MDC1 lacking the BRCT domain (Figure 7—figure supplement 1F). Taken together, these observations support a model in which MDC1 is constitutively tethered to chromatin by its PST repeat region and is enriched at DNA lesions via the interaction of the BRCT domains with γH2AX. Importantly, both the PST- and BRCT-domains are critical for efficient HR while the PST domain is not required for 53BP1 recruitment to DSBs, suggesting that while MDC1-chromatin binding is important for HR, this interaction may be dispensable for 53BP1-dependent DNA end-joining.

Figure 7 with 1 supplement see all
MDC1’s constitutive chromatin association is mediated by its PST repeat region.

(A) Graphical illustration of the primary sequence of MDC1 indicating the location of the PST repeat and BRCT domain and the associated deletion mutants generated to analyze effects on the MDC1-chromatin interaction. (B) SDS-PAGE gel of JF646-labeled cells depicting expression of transiently expressed WT, ΔPST, and ΔBRCT MDC1 in Halo-MDC1 knockout cells. (C) Representative images of transiently expressed, JF646-labeled MDC1 deletion mutants in living cells in the presence of absence of Zeocin. (D) Results of live-cell single-molecule analysis of transiently overexpressed MDC1 deletion mutants analyzed with single particle tracking and SpotOn. Each dot represents the indicated Dfree, Dbound, or Fraction Bound for MDC1 molecules appearing in at least three consecutive frames within a single cell. Red bar = median. Data are the combination of all analyzed cells imaged over three independent experiments (n≥60 cells total). Differently shaded points indicate data collected from separate biological replicates. Data were analyzed by two-way ANOVA with Tukey’s posthoc test. n.s.=not significant. ****=p < 0.0001.

Video 26
Representative live-cell single-molecule imaging movies of HaloTagged MDC1 deletion mutants transiently expressed in ΔMDC1 U2OS cells, labeled with JFX650, and acquired at 138 frames per second.

170x140 pixels with a pixel size of 0.16 μm.

Discussion

In this study, we have developed a panel of cell lines expressing HaloTagged DNA repair factors from their endogenous genomic loci. Using these cell lines, we have systematically analyzed the protein abundance, diffusion dynamics, and recruitment to DNA lesions of these factors, which are critical to maintain genome integrity in human cells. Our observations provide new insights into the molecular mechanism and kinetics of the recruitment of the shieldin complex to DNA lesions, which is a critical step in DSB repair pathway choice. In addition, our results reveal how the PST repeat and BRCT domains of MDC1 coordinate its chromatin binding and recruitment to DNA lesions to facilitate DSB repair. In total, our work is an important step towards developing a quantitative model of DNA repair in human cells and provides a number of new tools which will be tremendously useful for the DNA repair research community.

The shieldin complex is recruited to DNA lesions in distinct steps

The shieldin complex is essential to facilitate NHEJ downstream of 53BP1 by recruiting the CST complex and Polα/primase to DSBs (Mirman et al., 2018). While the pairwise protein and genetic interactions of the shieldin components are well understood (Gupta et al., 2018), how their dynamic recruitment to DSBs regulates repair via NHEJ was unknown. For instance, it is unclear to what extent shieldin is preassembled either as a complex (SHLD3-REV7-SHLD2-SHLD1), as subcomplexes (e.g. SHLD3-REV7 and SHLD2-SHLD1), or assembled entirely at a DSB. Some evidence supports the hypothesis of assembly at a break. Genetically SHLD3 is upstream of REV7, which in turn recruits SHLD2. SHLD2 directly associates with both single-stranded DNA and SHLD1, which recruits the CST complex to promote end fill-in (Gupta et al., 2018; Mirman et al., 2022; Noordermeer et al., 2018). Our live-cell single-molecule imaging demonstrates that SHLD1, SHLD2, and SHLD3 move through the nucleus with distinct diffusion coefficients. The diffusion coefficient of REV7 is comparable to SHLD2, but distinct from SHLD3. Observations by others have demonstrated that the interaction between SHLD3 and REV7 forms with extraordinarily slow kinetics which could be a key regulatory step in shieldin assembly at sites of DNA damage (Susvirkar and Faesen, 2022). Together our data strongly suggests that the shieldin complex components do not exist in a preassembled state, but rather associate at DNA lesions, which could allow the interaction of SHLD3 with REV7 to be rate limiting for shieldin complex formation.

Quantification of cellular protein abundance for HaloTagged SHLD1, SHLD2, and SHLD3 revealed that these proteins are expressed at comparable low levels. In addition, we observed both cytoplasmic and nuclear localization of each of these proteins by live-cell imaging. Therefore, the nuclear concentration for SHLD1, SHLD2, and SHLD3 is even lower. The low abundance and nuclear exclusion of SHLD1, 2, and 3 suggests that the recruitment of these factors to DSBs is tightly controlled. REV7 is largely found in the nucleus and 10-fold more abundant than SHLD3, which could significantly accelerate their slow complex formation observed in vitro (Susvirkar and Faesen, 2022). The low abundance of these proteins also has implications for understanding how these proteins find sites of DNA damage. For example, does recruitment of SHLD1, 2, and 3 occur by freely diffusing molecules encountering DSBs or could enrichment at DSBs be promoted by the establishment of 53BP1-mediated phase separated repair compartments (Kilic et al., 2019)? The low abundance of SHLD2 also raises the question of how it can compete for ssDNA binding with RPA which is thought to be more abundant and has a high affinity for ssDNA (Cho et al., 2022; Kim et al., 1992). One explanation could be the local enrichment of SHLD2 at DSBs mediated by 53BP1, comparable to POT1 recruitment to telomeric ssDNA overhangs by the shelterin complex (reviewed in Litman Flynn et al., 2012).

Interestingly, we observed large increases in the bound fraction of SHLD2 upon induction of DSBs, while there was no significant increase for SHLD1. This observation would be consistent with two potential models. First, SHLD1 may not bind every break where SHLD2 is present, but rather transiently visit breaks to facilitate Polα-dependent end fill-in via an interaction with CTC1 (Mirman et al., 2022). Another possible explanation is that SHLD1 only binds to one SHLD2 molecule at the 3’ DNA end, while SHLD2 may accumulate either in excess of its substrate (potentially accumulating in 53BP1 condensates), or by forming SHLD2 homopolymers on the ssDNA overhang.

It is well-documented that SHLD3 is the furthest upstream factor in the Shieldin complex coordinating 53BP1-RIF1 with REV7-SHLD2-SHLD1. In experiments overexpressing GFP-SHLD2, knockdown of endogenous RIF1, 53BP1, or SHLD3 led to a marked decrease in GFP-SHLD2 recruitment (Noordermeer et al., 2018), which we confirmed by knocking out SHLD3 in the Halo-SHLD2 cell line. Our results demonstrate that similar quantities of SHLD3 and SHLD2 are simultaneously recruited to DNA lesions, consistent with the established recruitment hierarchy. In addition, SHLD3 recruitment continuous after SHLD2 recruitment is saturated, which further supports the model that the shieldin complex is not pre-assembled in the nucleoplasm.

MDC1 and RIF1 are constitutive chromatin-binding proteins

Surprisingly, live-cell single-molecule imaging revealed constitutive chromatin-association of RIF1 and MDC1 in unperturbed cells. The fraction of chromatin bound MDC1 and RIF1 molecules is comparable to what we observe for Halo-H2B. While RIF1 and MDC1 are known to interact with chromatin, the absence of a substantial freely diffusing population leads to several questions for how these proteins function in DSB repair. First, if MDC1 and RIF1 are so immobile, how do they get recruited to DSBs? If the majority of MDC1 and RIF1 are always chromatin-associated, is the small pool of freely diffusing MDC1 and RIF1 sufficient for DSB repair or do chromatin bound molecules need to be locally reorganized or released from chromatin to form a focus? Additionally, RIF1 has many reported nuclear functions that require chromatin binding including telomere maintenance, 3D genome architecture, controlling replication timing, and DSB repair. It is unclear whether the static RIF1 that we observe contributes to one particular function or represents several distinct RIF1 activities. As such, our single-molecule imaging approach would be useful to test how disrupting RIF1 interactions by separation-of-function mutations alters its chromatin binding properties. For MDC1 we demonstrated that the constitutive chromatin association is mediated by its PST repeat domain which was recently reported to associate with the nucleosome acidic patch (Salguero et al., 2019). Some evidence suggests this domain is critical for promoting γH2AX-independent DSB repair, which would suggest that MDC1 constantly scans chromatin searching for DNA damage sites. While our observations would be consistent with that interpretation, it is also possible that MDC1 plays another important as yet unidentified role either in chromatin biology or specifically in DSB repair. Although rarely investigated, MDC1 is annotated as possessing 4 variants produced by alternative splicing, three of which we readily detect in roughly equal amounts. One of these variants has an exclusion of amino acids 1124–1410 constituting roughly 50% of the PST repeat region. Because endogenous genome editing preserves this normal splicing pattern, it is possible that this splice variant accounts for the freely diffusing MDC1 that we observe. Future studies will be required to address the role of the PST repeat region and the constitutive chromatin association of MDC1 in DNA repair.

Using single-molecule live cell imaging to analyze chromatin binding of DNA repair factors

We have developed a new method to analyze the recruitment of DNA repair factors to chromatin and DNA lesions using live cell single-molecule imaging. DNA repair factors that display a significant change in their chromatin bound fraction after DNA damage induction (e.g. DNA-PKcs, SHLD2, SHLD3, RNF169) can be studied using this approach. This approach is also amenable to performing residence time analysis of bound proteins which can be used to make meaningful insights into the affinity of DNA repair factors for their substrates in living cells. While we only tested zeocin induced DSBs in this study. It is possible that other means of inducing DNA lesions (e.g. PARP inhibitors, topoisomerase inhibitors) could result in significant changes in the chromatin bound fraction of other DNA repair factors. Importantly, due to our ability to detect individual molecules our approach is not limited to the analysis of DNA repair foci because it does not require the local accumulation of DNA repair factors to assess their chromatin binding. Previously, quantitative analysis of the recruitment of DNA repair factors to DNA lesions in living cells was limited to using laser microirradiation (LMI). LMI induces complex lesions, and the precise chemistry of the damaged DNA cannot be controlled. Our single-molecule imaging-based approach opens the door to quantitatively analyze the binding of DNA repair factors to DNA damage sites induced with a wide range of agents that lead to chemically well-defined DNA lesions. In addition, single-molecule imaging can directly measure kinetic parameters such as the dissociation rate of DNA repair factors from DNA damage sites, which will significantly advance our quantitative understanding of DNA repair in human cells.

In total, the results described in this study provide important insights into the protein abundance, diffusion dynamics, and recruitment kinetics to DNA breaks of a range of DNA repair factors. Our observations revealed that MDC1 and RIF1 constitutively associate with chromatin and demonstrate that the shieldin complex components are assembled in the context of DNA breaks but otherwise act independently. In addition, the collection of cell lines we have generated and the single-molecule imaging-based analysis of DNA repair factor recruitment to DNA breaks will be valuable tools for scientists studying DNA repair in human cells.

Materials and methods

Cell lines and cell culture

Request a detailed protocol

U2OS cells were obtained from the American Type Culture Collection (ATCC), were authenticated by STR profiling and tested negative for mycoplasma contamination. U2OS cells were cultured in RPMI containing 10% fetal bovine serum (FBS) and 100 units/mL penicillin and 100 μg/mL streptomycin in a humidified incubator maintained at 37 ° C with 5% CO2. For live-cell imaging experiments, cells were plated onto 24-well glass bottom plates and imaging was conducted using CO2-independent medium containing 10% FBS, 100 units/mL penicillin and 100 μg/mL streptomycin at 37 °C and 5% CO2.

Molecular cloning, plasmids, and genome editing

Request a detailed protocol

All gRNA, primer, and homology arm sequences used for cloning are included in Supplementary file 1. All gRNAs were cloned into BpiI-digested px330 backbone using standard procedures and ssDNA oligos were purchased from IDT. px330-U6-Chimeric_BB-CBh-hSpCas9 was a gift from Feng Zhang (Addgene plasmid #42230; http://n2t.net/addgene:42230; RRID:Addgene_42230). All homology-directed repair (HDR) donor plasmids were cloned using Gibson Assembly into pFastBac Dual backbone (Thermo Fisher, #10712024) linearized with HpaI. Gibson Assembly for each HDR donor consisted of three inserts including a left homology arm, right homology arm and an intervening sequence containing either the N-terminal or C-terminal HaloTag sequence. The N-terminal tag consists of a 3 x FLAG tag, an inverted SV40 promoter and Puromycin resistance cassette (PuroR) flanked by LoxP sites (to select for edited cells), the HaloTag, and a TEV protease cleavage site, followed by a short peptide linker. Conversely, the C-Terminal tag consisted of a short peptide linker and TEV protease cleavage site, followed by the HaloTag and the 3 X FLAG epitope, with the PuroR cassette oriented 3’ to the 3 X FLAG tag. This approach allowed for selectable direct insertion of the HaloTag at the C-Terminus of a protein without the need for the additional step of Cre-Lox recombination to remove the PuroR cassette. Homology arms consisted of >250 bp homologous to the genomic DNA directly upstream and downstream of the Cas9 cleavage site and were either ordered as double-stranded gene fragments from IDT or PCR amplified from genomic DNA. MDC1 deletion mutants were generated using rationally designed PCR primers to amplify MDC1 cDNA. Cloning of the Halo-MDC1 deletion mutants was done by Gibson Assembly into pRK2 (Modified from pHTN HaloTag CMV-neo; Promega; #G7721). All plasmids were confirmed for proper insertion/assembly by Sanger sequencing. Halo-H2B plasmid was kindly provided by Dr. Anders Hansen (Hansen et al., 2017) and the Halo-3xNLS plasmid was previously established and described (Klump et al., 2023).

Halo-MDC1 deletion mutants were expressed transiently by transfecting ~5 x 105 cells with 1 μg of plasmid DNA using FuGene 6 (Promega). For genome editing,~5 x 105 U2OS cells were transfected using FuGene 6 with either 500 ng or 1 μg of each gRNA/Cas9 and HDR plasmid in 6 well plates. Approximately 2–3 days post-transfection, edited cells were selected for with puromycin (1 μg/mL). After puromycin selection, cells were allowed to grow for ~2–3 weeks followed by sorting for single-cell clones. N-terminally edited cells were transfected with a plasmid encoding Cre to recombine out the PuroR cassette generating a 3xFLAG-HaloTagged protein. Cells were labeled with JF646-HaloTag ligand (JF646) and sorted based on JF646 signal (Grimm et al., 2015). For knockout of Halo-MDC1, Halo-53BP1, and SHLD3, cells were transfected with two gRNA plasmids. For Halo-MDC1 and Halo-53BP1 cells were isolated for single-cell clones by labeling with JF646 and sorting JF646-negative cells. Knockout of Halo-MDC1 and Halo-53BP1 was confirmed by loss of fluorescence by SDS-PAGE, Western blot, Sanger sequencing and for MDC1 with Inference of CRISPR Edits (ICE), while knockout of SHLD3 was confirmed by genomic PCR (Conant et al., 2022). Knock-in of the DR-GFP HR reporter into the AAVS1 locus was performed inΔMDC1 cells using AAVS1-DRGFP (Addgene plasmid #113193; RRID:Addgene_113193) and PX458-AAVS1 (Addgene plasmid #113194; RRID:Addgene_113194) which were kindly provided by Adam Karpf. After selection of edited cells with puromycin, cells were seeded by single-cell dilution in 96 well plates and clones were subsequently screened for DR-GFP insertion by genomic PCR.

SDS-PAGE and western blot

Request a detailed protocol

Mini-PROTEAN TGX stain-free gels (BioRad) were used for SDS-PAGE for most proteins except for MDC1, DNA-PKcs, and RIF1 where homemade 6% polyacrylamide gels were used. Protein samples were made by lysing cells with 2 x Laemmli buffer with β-mercaptoethanol. For in-gel fluorescence to detect HaloTagged proteins, cells were labeled with ~150–500 nM JF646 or JFX650 HaloTag ligand (JFX650; similar to JF646 but with improved photostability) for 30 min, washed with complete medium three times, allowed to rest for ~10 min, and subsequently washed twice with PBS before being lysed with buffer (Grimm et al., 2021). For in-gel fluorescence measurements of protein abundance, cells were plated in triplicate in 24-well plates, the following day samples were labeled with 500 nM JF646 for ~30 min, washed with PBS three times, lysed in 2 x Laemmli buffer with β-mercaptoethanol, and boiled for 5 min at 95 °C prior to gel loading. Standards were prepared using standards containing known femtomolar concentrations of FPLC-purified, JF646-labeled 3X-FLAG-HaloTag prepared in aliquots containing known cell concentrations, boiled for five minutes, and frozen at –80 °C until use. Fluorescence was detected using the Cy5.5 filter on a BioRad Chemidoc. Stain-free detection of protein loading was detected using the Stain-Free filter on a BioRad Chemidoc after 45 second UV activation. For protein abundance, the number of molecules were calculated by comparing the fluorescent signal for each protein relative to the standard curves for cell number and JF646-labeled 3X-FLAG-HaloTag. For TEV corrections,~120,000 cells were labeled with 500 nM JF646 for 30 min, washed once with PBS and harvested with 5 mM EDTA in PBS. Samples were lysed in 60 μL CHAPS lysis buffer and 20 μL incubated on ice with 5 units of TEV protease (New England Biolabs, P8112) for 30 min at which time 5 μL 6 x SDS sample buffer was added to 20 μL of the digested and undigested sample and boiled at 95 °C for 5 min. For TEV correction experiments where protease inhibitors used, samples were pre-treated with 10 μL of protease inhibitor cocktail (Sigma, P8340) for 30 min on ice followed by the addition of TEV protease. Comparisons between the JF646 signal ±digestion by TEV protease were performed by comparing the JF646 intensity for each sample after normalizing to stain-free protein signal. Transfers onto PVDF or Nitrocellulose membrane were conducted using either a Trans-Blot Turbo system (with Turbo transfer buffer) (BioRad) or by traditional wet tank transfer using CAPS Buffer with 10% Methanol (for DNA-PKcs, 53BP1, RIF1, and MDC1). Antibodies used were anti-FLAG-HRP (Sigma-Aldrich, A8592, RRID: AB_439702, 1:5000 dilution), anti-DNA-PKcs (a gift from Dr. Kathy Meek, 1:1000 dilution), anti-ATM (Santa Cruz, sc-135663, RRID: AB_2062962, 1:1000), anti-NBS1 (BioRad, VMA00403, 1:1000), anti-MDC1 (Novus Biologicals, NB100-395, RRID: AB_10001489, 1:1000), anti-RNF168 (GeneTex, GTX129617, RRID: AB_2886056, 1:1000), anti-53BP1 (Novus Biologicals, NB100-304, RRID: AB_10003037, 1:1000), anti-RIF1 (Bethyl Laboratories, A300-568A, RRID: AB_669806, 1:1000), anti-REV7 (Abcam, ab180579, RRID: AB_2890174, 1:1000), goat anti-mouse HRP (Invitrogen, 31430, 1:2000), goat anti-rabbit HRP (Invitrogen, 31460, 1:2000).

Purification of recombinant 3xFLAG-HaloTag protein

Request a detailed protocol

The plasmid encoding 6xHIS-3xFLAG-HaloTag protein (a kind gift from Dr. Daniel T. Youmans and Dr. Thomas R. Cech) was transformed into OneShot BL21(DE3) cells (Invitrogen). Cultures were grown to an OD600 of 0.6, induced with 1 mM IPTG, and grown at 18 °C for 16 hr shaking at 180 RPM. Bacterial cells were harvested by centrifugation at 5000xg for 15 min at 4 °C and frozen at –80 °C for 1 hr. Frozen cell pellets were resuspended in lysis and wash buffer (50 mM sodium phosphate buffer pH 8.0, 300 mM sodium chloride, 10 mM imidazole, 5 mM beta-mercaptoethanol). Cells were lysed by addition of 0.5 mg/ml lysozyme and sonication (40% amplitude, 90 s of sonication, 10 s pulses, 20 s pause, Fisherbrand Model 505, 0.5 inch tip) in an ice water bath. Cell lysates were cleared at 40,000xg and 4 °C for 30 min and incubated with fast flow nickel Sepharose (Cytiva) for 1 hr at 4 °C. The resin was washed three times with lysis and wash buffer and the 6xHIS-3xFLAG-HaloTag was eluted in elution buffer (50 mM sodium phosphate buffer pH 7.0, 300 mM sodium chloride, 250 mM imidazole, 5 mM beta-mercaptoethanol). The 6xHIS-3xFLAG-HaloTag was further purified using size exclusion chromatography using a superdex 75 column into 50 mM Tris pH 7.5, 150 mM potassium chloride, 1 mM DTT. Peak fractions were combined, concentrated, supplemented with 50% glycerol, snap frozen in liquid nitrogen and stored at –80 °C. To fluorescently label the 6xHIS-3xFLAG-HaloTag, we incubated the protein with a twofold excess of JF646 HaloTag-ligand overnight at room temperature. Excess fluorescent dye was removed by size exclusion chromatography using a superdex 75 column into 50 mM Tris pH 7.5, 150 mM potassium chloride, 1 mM DTT. The protein concentration and labeling efficiency was determined by absorption spectroscopy using ε280nm = 41,060 M–1 cm–1 for the 6xHIS-3xFLAG-HaloTag (calculated using primary protein sequence and the ExPASy ProtParam tool) and ε646nm = 152,000 M–1 cm–1 for the JF646 fluorescent dye (Grimm et al., 2015).

Clonogenic survival assays

Request a detailed protocol

On the day prior to treatment cells were seeded in triplicate at a density of 500 cells per well in six-well plates. The following day, cells were treated with Zeocin (Gibco) at the indicated concentrations for four hours in complete medium. After treatment, medium was removed and replaced with fresh complete medium. For assays including the ATM inhibitor (ATMi) (KU-55933; Selleckchem), cells were pre-treated for 2 hr, during Zeocin treatment, and for 24 hr post-Zeocin with 10 μM ATMi. Cells were allowed to grow until colonies reached a size of >50 cells (approximately 7–10 days) at which time media was removed, wells washed with PBS and cells fixed and stained in crystal violet solution (20% ethanol and 1% w/v crystal violet). After incubation excess staining solution was removed and plates gently washed in diH2O and air-dried. Plates were imaged on a BioRad Chemidoc using the Coomassie Blue filter set. Colony counts were determined using ImageQuant TL 8.2. All colony survival assays were performed in triplicate at least three times and are plotted as the average ± S.D.

Flow cytometry

Request a detailed protocol

Four independent experiments were performed to quantify relative protein abundance using flow cytometry. On the day prior to performing flow cytometry,~100,000 cells of each HaloTagged protein clone were seeded into 24-well plates. The next day, cells were labeled with 500 nM JF646 for ~30 min, washed three times with PBS, and allowed to rest for 5 min in complete medium to allow unbound dye to leak out of the cells. Cells were harvested using 5 mM EDTA in PBS, fixed in 2% paraformaldehyde for 10 min, washed once with PBS and resuspended in PBS with 1% bovine serum albumin. Samples were run on a Cytek Aurora spectral flow cytometer and >7500 events were collected per protein for each replicate to determine mean fluorescence intensity. Data were analyzed in FCS Express 7. For DR-GFP experiments, ~200,000 cells were nucleofected in RPMI plus 50 mM HEPES with 400 ng I-SceI plasmid and 600 ng of each HaloTagged MDC1 or 3XFLAG-Halo-NLS plasmid and seeded in six-well plates. Forty-eight hr post-nucleofection, cells were labeled with 150 nM JFX650, collected by trypsinization, and samples were run on a BD Accuri C6 cytometer and >5000 JFX650-positive cells were collected for each sample to determine the percentage of GFP-positive cells. Data were analyzed in FCS Express 7.

Laser microirradiation

Request a detailed protocol

Laser microirradiation (LMI) was carried out on a Olympus IX83 inverted microscope equipped with a 4-line cellTIRF illuminator (405 nm, 488 nm, 561 nm, 640 nm lasers), an Excelitas X-Cite TURBO LED light source, a Olympus UAPO 100 x TIRF objective (1.49 NA), a CAIRN TwinCam beamsplitter, 2 Andor iXon 897 Ultra EMCCD cameras, a cellFRAP with a 100 mW 405 nm laser, an environmental control enclosure and operated using the Olympus cellSense software. Cells were seeded the day before LMI onto a 24-well glass bottom plate. On the day of the experiment, cells were labeled with 150 nM JFX650 for 30 min, washed three times with complete media and presensitized with Hoechst (1 μg/mL) for 10 min before washing and adding fresh complete medium. Plates were placed on the microscope stage which was pre-warmed to 37 °C with 5% CO2. Cells were irradiated using a 20ms pulse at 25% laser power using drawn interpolated lines or a diffraction limited spot (only used for REV7 due to diffuse staining pattern which made it difficult to quantify over time). Images were acquired using a 100 x objective and fluorescence imaged with excitement by the 630 nm LED light source at various rates (e.g. every 1 s, 2 s, 5 s, or 10 s). To quantify the fluorescent signal after LMI, we converted movies to.tif files. Irradiated cells were cropped and images drift corrected using NanoJ in Fiji (Laine et al., 2019). Using drift corrected movies, an ROI was placed around the irradiated area and we measured the mean intensity within the ROI over time. Baseline fluorescence was subtracted from each sample in order to calculate the relative increase in fluorescence that accumulated over time. To normalize fluorescence intensity values, each cell had its highest intensity frame set to one in order to normalize fluorescence intensity values between samples. For averaging normalized intensity values for all cells, the frame number with the highest average intensity was set to one. Data were plotted in GraphPad Prism and recruitment half-time was determined by fitting with an exponential one-phase association model (Y=Y0 + (Plateau – Y0)*(1-exp(-K*x))) where Y0 was set to zero and Plateau was set to one.

Live-cell imaging

Request a detailed protocol

Live-cell and live-cell single-molecule imaging was performed on the same microscope described above. For live-cell imaging of HaloTagged DDR protein localization and DNA-damage induced foci, samples were densely labeled with 500 nm JF646 for 15–30 min, washed three times with complete medium and allowed to rest for 10 min before adding CO2-independent medium. For Zeocin-treated samples, cells were treated for one hour prior to HaloTag labeling with 100 μg/mL Zeocin. Z-stack images were acquired at 37 °C in the presence of 5% CO2 with a 100 x objective and the 640 nm laser. These experiments were performed twice imaging >20 cells per experiment. For live-cell single-molecule imaging, HaloTagged DDR proteins were labeled with JFX650 at differing concentrations to achieve single-molecule density (0.1 nM for 30 s up to 20 nM for 1 min). After labeling, cells were washed three times with complete medium and allowed to rest for 10 min before adding CO2-independent medium. For Zeocin-treated samples, cells were treated for one hour with Zeocin (100 μg/mL) prior to protein labeling with JFX650. Imaging was performed at 37 °C and 5% CO2 with the 100 x objective and the 640 nm laser with highly inclined laminated optical sheet (HILO) illumination (light angled between 1.29 and 1.32 depending on the sample). Images were acquired at 138 fps for 3000 frames followed by a brightfield image to visualize the cell. For dual color imaging of Halo-53BP1, cells were first labeled with 2.5 nM JFX650 for one minute followed by 100 nM JF503 for 10 minutes. Images were acquired at 202 fps and collected using a 60 x objective and the 640 nm and 488 nm laser lines. Imaging of MDC1 mutants in the presence of absence of Zeocin were acquired at 202 fps with a 60 x objective and the 640 nm laser line with HILO illumination.

Analysis of live-cell single-molecule imaging

Request a detailed protocol

Single-particle tracking (SPT) was performed in MATLAB 2019a using a version of SLIMfast allowing for analysis of TIFF files (Hansen et al., 2018). Settings for SPT used in the analysis were as follows: Exposure Time = 7.24ms (4.95ms for data presented in Figure 7—figure supplement 1C), NA = 1.49, Pixel Size = 0.16 µm (0.1083 for data presented in Figure 7—figure supplement 1C), Emission Wavelength = 664 nm, Dmax = 5 µm2/s, Number of gaps allowed = 2, Localization Error = –5, Deflation Loops = 0. While most HaloTagged DDR proteins exhibited near complete nuclear localization, REV7, SHLD3, SHLD2, and SHLD1 possessed mixed cytoplasmic and nuclear localizing fractions. For these proteins, the brightfield image was used to generate a nuclear mask in FIJI to separate cytoplasmic from nuclear tracks. Only nuclear tracks were used for analysis of protein diffusion in SpotOn. SPT files were then used in SpotOn in MATLAB to determine diffusion coefficients and the percentage of bound versus free particles. The following settings were used in SpotOn analysis of SPT files: TimeGap = 7.24ms (4.95ms for data presented in Figure 7—figure supplement 1C), dZ = 0.700 µm, GapsAllowed = 2, TimePoints = 8, JumpsToConsider = 4, BinWidth = 0.01 µm, PDF-fitting, D_Free_2State = [0.5 25], D_Bound_2State = [0.0001 0.5]. All live-cell single molecule imaging experiments were performed three times acquiring >20 cells per condition per experiment. Comparisons of Diffusion coefficients and fractions bound were performed in GraphPad Prism by two-way ANOVA with Tukey’s posthoc test. To filter out trajectories that overlapped with DNA repair foci, we generated a mask using the quantitatively labeled Halo-53BP1 signal and assigned Halo-53BP1 (JFX650) trajectories whose coordinates overlapped with the mask for at least on frame to the DNA repair foci group.

Residence time analysis

Request a detailed protocol

For the residence analysis of low mobility particles, we averaged the intensity of each pixel of single-molecule live cell imaging movies over 10 consecutive frames reducing the effective imaging rate from 138 to 13.8 frames per second. This approach blurs out the signal from mobile molecule because they do not reside in the same location for multiple consecutive frames. Single particle tracking was then carried out constraining Dmax to 0.5 µm2/s to exclusively track low mobility molecules, which are assumed to be chromatin bound. Track length distributions were plotted as survival probabilities (1 – Cumulative density function of the track lengths). To determine the rate constants and corresponding half-lives the decay functions were fit two a two-component exponential decay function.

Immunofluorescence

Request a detailed protocol

Two days prior to performing immunofluorescence, ~1 million ΔMDC1 cells were nucleofected with two mg of Halo-MDC1 WT, ΔPST, or ΔBRCT plasmid using a Lonza 4D nucleofector. The following day,~250,000 cells were seeded onto glass coverslips in six-well plates in complete medium. The next day, cells were left untreated or treated with 10 mg/mL Zeocin for one hour after which samples were labeled with 150 nM JF646 HaloTag ligand for 10 min, washed three times, and incubated in complete medium for 10 min to allow unbound ligand to leak out of the cells. Next, cells were washed with PBS and fixed in 4% formaldehyde prepared in PBS for 10 min at room temperature. After fixation, coverslips were washed with PBS and cells permeabilized with 0.2% Triton-X 100 for 2 min. Next, coverslips were washed twice with antibody dilution buffer (ABDIL) (3% BSA in PBS-Tween 20 (PBS-T)) and incubated in ABDIL for one hour. Coverslips were incubated with anti-53BP1 primary antibody (Cell Signaling Technology Cat# 4937, RRID:AB_10694558) in ABDIL at 1:200 dilution for 1 hr, followed by 35-min washes with PBS-T, and 1 hr incubation with goat anti-rabbit antibody conjugated to AF488 (Thermo Fisher Scientific Cat# A-11034, RRID:AB_2576217) at 1:500 dilution for 1 hr. After washing three times with PBS-T, DNA was stained with Hoechst, coverslips mounted onto slides with ProLong Diamond Antifade Mountant and sealed with nail polish. Samples were imaged using a DeltaVision Elite 642 microscope with a 60 x PlanApo objective (1.42 NA) and a pco.edge sCMOS camera. Images were processed by deconvolution in DeltaVision Softworx followed by maximum intensity projection in ImageJ.

Data availability

All uncropped images for gels and blots included in this manuscript have been provided as source data.

References

Decision letter

  1. Wolf-Dietrich Heyer
    Reviewing Editor; University of California, Davis, United States
  2. Detlef Weigel
    Senior Editor; Max Planck Institute for Biology Tübingen, Germany
  3. Judith Miné-Hattab
    Reviewer; Institut Curie, PSL University, Sorbonne Université, CNRS, Nuclear Dynamics, France

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting the paper "Systematic analysis of the molecular and biophysical properties of key DNA damage response factors" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Wolf-Dietrich Heyer as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by a Senior Editor. The following individuals involved in the review of your submission have agreed to reveal their identity: Markus Löbrich (Reviewer #2); Judith Miné-Hattab (Reviewer #3).

Comments to the Authors:

We are sorry to say that, after consultation with the reviewers, we have decided that this work will not be considered further for publication by eLife.

The reviewers and Reviewing Editor recognize the potential utility of the reported tools for the scientific community and appreciate that these tools can provide certain novel insights. However, the analysis is incomplete in several areas and the conclusions seem insufficiently supported by the experimental evidence. The potential revisions would be extensive and consume more time than compatible with eLife's editorial policy.

If you decide to address the extensive revisions requested, we would encourage resubmission as a new manuscript, and we would make an effort to recruit the same reviewers to assess the work, which would be treated as a new submission. The major issues are the functionality of the tagged proteins, especially MDC1 and SHLD1/2 for which major conclusions are reached (#3), completeness of the single particle tracking analysis (#5), and orthogonal genetic validation of the major conclusions from the kinetic analysis (#11, 13, 16).

Life cell imaging provides unprecedented insights into cellular processes, and advances in fluorescence and microscopy allow the identification and tracking of single protein particles in four dimensions. The manuscript reports the creation of a set of useful cell lines with Halotag fusions to 12 key proteins acting in the DNA damage response, namely ATM, NBS1, MDC1, RNF168, RNF169, 53BP1, RIF1, REV7, SHLD1/2/3, and DNA-PKcs. The fusions were carefully validated molecularly and functionally, leading to detectable expression of Halotagged proteins in protein gels. All proteins, with the exception of ATM-Halo, showed the expected cellular localization in undamaged cells and led to an increase in focus formation in response to DNA damage (Zeocin) with the exceptions of ATM and DNA PKcs. Clonogenic survival assays demonstrated the functionality of the fusion proteins, with the exception of Halo-ATM, which appeared similar to a loss of function for both a C- and an N-terminal fusion, and Halo-MDC1, Halo-SHLD1/2, and Halo-53BP1, which showed partial loss of function. Using flow cytometric and in-gel approaches, the steady protein level for all Halo fusions was determined with a good correlation between both methods. All experiments were conducted with two independent clones for each fusion (except NBS, SLD2, 53BP1). All cell lines were homozygous for the Halo tag, with the exception of Halo-SHLD2, where one allele was tagged and the other was a frameshift allele.

Kinetic recruitment experiments using the Halo-tagged proteins were conducted using laser microirradiation-induced DNA damage. The data for components of the Shieldin complex showed significantly different recruitment kinetics for SHLD2 and 3, suggesting that this protein complex assembles at the site of DNA damage rather than being pre-assembled. The kinetic resolution for the other proteins allowed the identification of additional differences in protein recruitment at laser-induced DNA damage. The caveat working with non-physiological laser-induced DNA damage could have been considered and potentially be selectively complemented with orthogonal ways to induce DNA damage such as Cas9-mediated DSBs.

Single particle tracking was used to determine the nuclear diffusion of the single proteins in the presence and absence of DNA damage (zeocin). It is unclear if the technology allows the tracking of a single molecule. The results identified significant differences between the Shieldin subunits 1, 2, and 3, corroborating the conclusion that they do not exist as a pre-assembled complex. The data for MDC1 and RIF1 suggest that both are largely chromatin-associated in undamaged cells, and follow-up experiments show that the MDC1 PST domain is responsible for this, whereas the BRCT domain of MDC1 is critical for DNA damage-induced chromatin association, as previously shown. The analysis of the MDC1 domains lacks experiments under DNA damage conditions (zeocin).

Overall, this manuscript is well-written and documented but the analysis remains incomplete in several instances.

Recommendations for the authors:

1) The ATRX mutation in U2OS cells affects DNA repair pathway choice (PMID: 29937341, PMID: 33431668). This caveat should be considered and discussed.

2) I do not understand the comment in lines 210/211: "Importantly, the differences in absolute protein number between independent genome-edited clones could be the consequence of a different number of alleles being modified with the HaloTag." It is stated in lines 115-117 that all lines are homozygously tagged except for one, where the other allele is a frameshift likely resulting in expressing an unstable truncation protein. There should be no variation in tagged alleles, or am I missing something?

3) The survival of cells expressing Halo-tagged version of ATM, MDC1, 53BP1, SHLD1, and SHLD2 is reduced after zeocin treatment compared to wild-type cells. Thus, the functionality of these tagged proteins is questionable. In particular, the analysis focuses on MDC1 although MDC1 is one of the less functional tagged proteins. For example, MDC1 is reported as less mobile than histones H2B: is this really possible? Could it be an effect of the partial loss of MDC1 functionality? If not, how can such slow mobility be explained? The manuscript also reports that the constitutive interaction between MDC1 and chromatin is mediated by the PST repeat domain of MDC1: again, it is necessary to be careful about this conclusion since the functionality is reduced in the Halo-tagged version.

We appreciate that fully functional fusions may be out of reach, but the limitations need to be acknowledged and discussed. Have alternative tag designs been tried?

4) In the section: Functional validation of HaloTagged DDR proteins:

In the absence of zeocin treatment, cells expressing Halo-MDC1 exhibit many spontaneous foci. Is it something known and if not, how can that be explained?

Cells expressing ATM-Halo do not form foci after zeocin treatment: please comment.

Concerning DNA-PK: this protein is highly abundant in the cell; however, it does not form foci after zeocin treatment. Is there an explanation? does it mean that even if the protein is very abundant, very few DNA-PK molecules are present within foci and sufficient for the next steps of NHEJ? Does it form a visible line after micro-irradiation?

5) Page 12, line 206: It is stated that MDC1 has a higher protein abundance than ATM, SHLD1, SHLD2, and SHLD3 but one of the two MDC1 clones analyzed has the lowest protein abundance of all clones analyzed in this study (2400 molecules per cell according to the text and table I). Why are the two MDC1 clones differ so drastically from each other? Confusingly, the big difference between the two clones is not seen in the bar chart in Figure 3B and is not measured by flow cytometry.

6) Page 12, line 231: It is stated that the adjustment factors for DNA-PKcs are 0.62 and 0.79 but the application of these factors increases the molecule number for clone 1 but decreases it for clone 2 (the latter seems wrong).

7) Single Particle Tracking analysis:

Using analysis of single particle tracking, the authors can measure the diffusive properties of repair proteins. Diffusion is estimated in the presence and in absence of zeocin treatment. Thus, the cells contain many foci: all the traces in the nucleus are analyzed at once, inside and outside foci. The authors then used a 2 population model to fit the distribution of protein displacements.

The SPT analysis allows the authors to give some interesting mechanistic insight but the authors could extract much more information from their data. Why are the values of Dslow not provided? The authors interpret the slow population as bound to damaged DNA. However, it is known that some proteins diffuse relatively fast inside repair foci, especially if they are able to form liquid-liquid phase separation. In addition, some proteins might diffuse slowly outside of foci, because of non-specific interactions (or even for MDC1 for example). Thus, it is possible that the slow molecules are a mixture of the molecules inside the foci with molecules exhibiting chromatin binding outside of the repair foci.

There is no visualization of the traces allowing us to see if the slow molecules are indeed inside foci and the fast ones are outside. It would be essential to be able to see this for at least 1 cell for each repair protein.

Are mixed traces observed, with a slow and a fast part?

Is it possible to estimate the residence time of each protein inside repair foci or on their substrate?

Are the proteins inside foci exchanging with the rest of the nucleus or are they stuck inside the focus during the entire trace?

Since you use bright JF, it should be possible to have long traces: the authors should show a distribution of the traces' length for each repair protein.

Do you see a change in protein diffusion in the absence of zeocin treatment and in the presence of zeocin treatment outside of foci?

The authors also use H2B and NLS as controls. The values obtained should be compared with values found in the literature.

Finally, it is not clear how the histones H2B are tagged in this study: is it also an endogenous H2B-Halo tag? Is it a stable cell line but not endogenous, or is it a transient transfection?

8) Laser micro-irradiation induces massive damage and may not be reflective of physiological encountered DNA damage. Have the authored considered using Cas9-induced DSBs as a defined and targeted DNA damage? I understand that adding such experiments for all proteins would be a massive endeavor, but maybe this could be done for MDC1 and/or the Shieldin complex. Regardless, the limitations of the laser micro-irradiation approach should be discussed.

9) Figure 6 lacks data for the analysis in the presence of zeocin, in the way it was done in the analysis for Figure 5. Such data will corroborate the foci analysis and potentially reveal differences in the recruitment of MDC1 to damaged sites.

10) What is the evidence that a single molecule can be tracked in Figures 5 and 6, as opposed to a single particle that may be composed of multiple proteins?

11) Page 13, line 256: It was surprisingly observed that REV7 and SHLD3 have vastly different recruitment times to laser-induced damage although both factors are known to interact. At the end of the Discussion section, it is speculated that this might reflect REV7's role in TLS but the reader would benefit if such an interpretation was offered earlier in the paper. Moreover, if this interpretation was correct, one would expect that REV7 was recruited to laser damage independently of SHLD3. This should be tested by siRNA-mediated depletion of SHLD3 (or other factors upstream of SHLD3).

12) Page 14, lines 260-263: It is observed that SHLD1 is not recruited to laser damage although it forms foci after zeocin treatment. The authors then speculate that this might suggest that SHLD1 recruitment is the key regulatory step for a fill-in reaction. This is not clear. Are the authors suggesting that fill-in takes place at zeocin-induced breaks but not at laser-induced breaks? Please clarify.

13) Page 14, line 266: The authors state that the measured recruitment kinetics provide insight into the interdependencies of the shieldin complex components. As evidenced by the very early recruitment kinetics of REV7 (possibly due to its role in TLS), this may not be true since the factors could have roles in other processes. Thus, the evaluation of interdependencies requires the measurement of the recruitment kinetics in situations when individual components of the shieldin complex are depleted by siRNA technology. Thus, the authors should assess the recruitment kinetics in cells depleted for SHLD2/3, REV7, or 53BP1/RIF1.

14) Page 17, line 319: SHLD1 is not recruited to chromatin after DSB induction despite its low abundance, and the authors state that this is consistent with its lack of recruitment to laser damage. However, it is clearly shown to form foci at DSB. Moreover, how does this finding fit the author's suggestion from above that SHLD1 is the key regulator of the fill-in reaction?

15) The authors try to address the unexpected findings for SHLD1 (no recruitment to laser damage and no increase in the chromatin-bound fraction after zeocin treatment) in the discussion on page 22, last paragraph, and suggest that SHLD1 may either not bind every DSB where SHLD2 is present or may reside at a DSB in lower copy numbers than SHLD2. Both explanations appear inconsistent with the robust formation of SHLD1 foci at zeocin-induced DSB, the first suggestion can easily be tested with co-localization experiments using HaloTag SHLD1 and SHLD2 ab or vice versa.

16) Page 21, line 390: The authors suggest the model that RNF169 delays 53BP1 recruitment to DSBs. Although this interpretation is consistent with the presented data, the model should be tested by 53BP1 recruitment measurements after siRNA-mediated depletion of RNF169.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Systematic analysis of the molecular and biophysical properties of key DNA damage response factors" for further consideration by eLife. Your revised article has been evaluated by Detlef Weigel (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

Life cell imaging provides unprecedented insights into cellular processes, and advances in fluorescence and microscopy allow the identification and tracking of single protein particles in four dimensions. The manuscript reports the creation of a set of useful cell lines with Halotag fusions to 12 key genes acting in the DNA damage response, namely ATM, NBS1, MDC1, RNF168, RNF169, 53BP1, RIF1, REV7, SHLD1/2/3, and DNA-PKcs. The fusions were carefully validated molecularly and functionally, leading to detectable expression of Halotagged proteins in protein gels. All proteins, with the exception of ATM-Halo, showed the expected cellular localization in undamaged cells and led to an increase in focus formation in response to DNA damage (Zeocin) with the exceptions of ATM and DNA PKcs. Clonogenic survival assays demonstrated the functionality of the fusion proteins, with the exception of Halo-ATM, which appeared similar to a loss of function for both a C- and an N-terminal fusion, and Halo-MDC1, Halo-SHLD1/2, and Halo-53BP1, which showed partial loss of function. Using flowcytometric and in gel approaches, the steady protein level for all Halo fusions was determined with good correlation between both methods. All experiments were conducted with two independent clones for each fusion (except NBS, SLD2, 53BP1). All cell lines were homozygous for the Halo tag, with the exception of Halo-SHLD2, where one allele was tagged and the other was a frameshift allele.

Kinetic recruitment experiments using the Halo-tagged proteins were conducting using laser microirradiation induced DNA damage. The data for components of the Shieldin complex showed significantly different recruitment kinetics for SHLD2 and 3, suggesting that this protein complex assembles at the site of DNA damage rather than being pre-assembled. The kinetic resolution for the other proteins allowed identification of additional differences in protein recruitment at laser-induced DNA damage. The caveat working with non-physiological laser-induced DNA damage could have been considered and potentially be selectively complemented with orthogonal ways to induce DNA damage such as Cas9-mediated DSBs.

Single particle tracking was used to determine the nuclear diffusion of the single proteins in the presence and absence of DNA damage (zeocine). It is unclear if the technology allows tracking of a single molecule. The results identified significant differences between the Shieldin subunits 1, 2, and 3, corroborating the conclusion that they do not exist as a preassembled complex. The data for MDC1 and RIF1 suggest that both are largely chromatin-associated in undamaged cells, and follow-up experiments show that the MDC1 PST domain is responsible for this, whereas the BRCT domain of MDC1 is critical for DNA damage-induced chromatin association, as previously shown. The analysis of the MDC1 domains lacks experiments under DNA damage conditions (zeocine).

The two main conclusions from the work are that (1) the individual subunits of the Shieldin complex are recruited independently to sites of DNA damage, and (2) MDC1 and RIF1 are bound constitutively to chromatin. Although these conclusions are supported by the data, some inconsistencies with the literature remain unresolved. For example, the authors report that SHLD2 is recruited to laser tracks before SHLD3, while previous work demonstrated a genetic requirement for SHLD3 to recruit SHLD2 to sites of DNA damage. Further, it remains an open question how and when MDC1 and RIF1 are recruited to sites of DNA damage if they are constitutively bound to chromatin.

The use of Halo ligands with different emission spectra to simultaneously monitor single-particles and DNA repair foci is very elegant and can potentially be used to distinguish the behavior of different subpopulations of a DNA repair factor.

The manuscript is well-written, but some literature reference and information in the methods section are missing.

In conclusion, this is an interesting study that applies novel microscopy techniques to DNA double-strand break repair proteins. However, the study remains somewhat descriptive, which limits the mechanistic insight gained from the study and the analysis remains incomplete in several instances for lack of genetic corroboration of the main conclusion about the Shieldin complex recruitment.

Recommendations for the authors

Essential revisions

1) Line 144: The authors conclude that "the HaloTag does not impact the proper cellular localization of these proteins" based on fluorescence microscopy of the Halo-tagged proteins after JF646 labeling. This conclusion cannot be made, because it would require examination of the localization of the untagged proteins under the same conditions. Please qualify your statement.

2) Line 146: The authors conclude that "HaloTagging ATM at the N-terminus led to nuclear exclusion". The authors cannot make this conclusion without imaging the untagged ATM under the same condition. Please qualify your statement.

3) Line 183: It cannot be concluded that "most possess full DNA repair functionality" unless the cell lines expressing Halo-tagged proteins are compared to their gene knockout counterparts. This was only performed for 53BP1 and MDC1, where partial functionality was observed. Please qualify your statement.

4) Page 13, line 256: It was surprisingly observed that REV7 and SHLD3 have vastly different recruitment times to laser-induced damage although both factors are known to interact. At the end of the Discussion section, it is speculated that this might reflect REV7's role in TLS but the reader would benefit if such an interpretation was offered earlier in the paper. Moreover, if this interpretation was correct, one would expect that REV7 was recruited to laser damage independently of SHLD3. This should have been tested by siRNA-mediated depletion of SHLD3 (or other factors upstream of SHLD3) and this limitation should be explicitly acknowledged in the text.

5) Page 14, line 266: The authors state that the measured recruitment kinetics provide insight into the interdependencies of the shieldin complex components. As evidenced by the very early recruitment kinetics of REV7 (possibly due to its role in TLS), this may not be true since the factors could have roles in other processes. Thus, the evaluation of interdependencies requires the measurement of the recruitment kinetics in situation when individual components of the shieldin complex are depleted by siRNA technology. Thus, the authors should assess the recruitment kinetics in cells depleted for SHLD2/3, REV7 or 53BP1/RIF1. The caveat of roles in independent processes should be explicitly mentioned.

6) Line 293: It is counter-intuitive that SHLD2 foci depends on SHLD3, which is recruited to foci significantly later than SHLD2. The two-step model suggested in the Discussion to explain this observation is not supported well by the data, because one would expect the initial (pre-SHLD2) step of SHLD3 to also be detected in the LMI experiments. Further, SHLD2 actually dissociates when SHLD3 associates with laser stripes, which is not discussed by the authors. It would strengthen the manuscript if the model could be tested experimentally by the authors.

7) A similar study was conducted previously (Aleksandrov et al. 2018), which reported recruitment half-times to laser stripes significantly different than the current study for some proteins and in other cases similar half-times. For example, the recruitment times for MDC1, RNF168, RNF169, and 53BP1 was reported by Aleksandrov to be 35s, 78s, 203s, and 307s, where the current study report 77s, 69s, 186s, 669s. It would be in place to reference the previous study and compare findings.

8) It is surprising that Halo-H2B only displays 66% chromatin binding. The assumption is that JF646 binds irreversibly to the Halo-tag, but is it possible that some free JF646 is present and gives rise to the "free" pool of fluorophores? This could easily be tested by formaldehyde fixation which should give 100% chromatin binding if no free JF646 is present.

9) Line 331: A two-state diffusion model is assumed where particles either freely diffuse or are chromatin bound. How would the conclusions be affected if a third state was allowed where a protein is part of a slow diffusing macromolecular complex.

10) Figure 5: The scatter plots should be displayed as "super plots" where data points for each of the 3-4 independent experiments are presented in different colors/symbols (Lord et al. 2020). This would reveal any systematic differences between experiments. For example, it looks like RNF169 data points can be divided into two populations.

11) Page 21, line 390: The authors suggest the model that RNF169 delays 53BP1 recruitment to DSBs. Although this interpretation is consistent with the presented data, the model could be tested by 53BP1 recruitment measurements after siRNA-mediated depletion of RNF169. This limitation should be explicitly acknowledged in the text.

12) The sources of several plasmids are missing e.g. pX330 in line 665.

13) A table with oligonucleotides and gRNAs used in the study should be included.

14) Line 511: Which data allow the authors to conclude that the PST domain of MDC1 "facilitates DDR signal amplification"?

15) The authors conclude that MDC1 and RIF1 are constitutively associated with chromatin. If this is the case then one might expect the localization of MDC1 and RIF1 to follow that of condensed chromosome when cells progress from interphase into mitosis. Indeed, there is evidence for this for RIF1 (Watts et al. 2020), but for MDC1 I could not find such evidence in the literature. However, I suspect that the authors in their data have images of mitotic cells that could answer this question.

16) To quantify the mobility inside foci versus outside foci, and the flux of proteins in and out of foci, would it be possible to make a density map of all the repair proteins in the nucleus? Using this density map, the foci appear clearly, and it is then possible to separate trajectories within foci from those outside foci.

17) Figure S5B: After zeocin treatment: is it possible that the lines between foci are misslinking?

References

Aleksandrov R, Dotchev A, Poser I, Krastev D, Georgiev G, Panova G, Babukov Y, Danovski G, Dyankova T, Hubatsch L et al. 2018. Protein Dynamics in Complex DNA Lesions. Mol Cell 69: 1046-1061 e1045.

Lord SJ, Velle KB, Mullins RD, Fritz-Laylin LK. 2020. SuperPlots: Communicating reproducibility and variability in cell biology. J Cell Biol 219.

Watts LP, Natsume T, Saito Y, Garzon J, Dong Q, Boteva L, Gilbert N, Kanemaki MT, Hiraga SI, Donaldson AD. 2020. The RIF1-long splice variant promotes G1 phase 53BP1 nuclear bodies to protect against replication stress. eLife 9.

https://doi.org/10.7554/eLife.87086.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Essential revisions:

1) The ATRX mutation in U2OS cells affects DNA repair pathway choice (PMID: 29937341, PMID: 33431668). This caveat should be considered and discussed.

We thank the reviewers for the comment regarding ATRX. The two references supplied do demonstrate that ATRX is not expressed in U2OS cells, however the Cancer Cell Line Encyclopedia characterization of human cancer cell lines using next-generation sequencing did not detect an ATRX mutation in U2OS cells, so the absence of expression may be due to another underlying cause (PMID: 31068700). We have now included a statement regarding the absence of ATRX expression in U2OS cells which plays important roles in DSB repair.

2) I do not understand the comment in lines 210/211: "Importantly, the differences in absolute protein number between independent genome-edited clones could be the consequence of a different number of alleles being modified with the HaloTag." It is stated in lines 115-117 that all lines are homozygously tagged except for one, where the other allele is a frameshift likely resulting in expressing an unstable truncation protein. There should be no variation in tagged alleles, or am I missing something?

While genomic PCR in Figure 1 demonstrates that all alleles are homozygously edited, these data do not report on how many alleles are present within each clone. Because cancer cell lines can possess inherent variations in karyotype within a cell population, we can not exclude the possibility that differences in relative expression of tagged proteins could be either a consequence of either a different number of alleles present in the clone that was genetically modified or a result of inherent clonal variations in expressing the protein of interest. We have now included a statement clarifying these two possibilities in the text.

3) The survival of cells expressing Halo-tagged version of ATM, MDC1, 53BP1, SHLD1, and SHLD2 is reduced after zeocin treatment compared to wild-type cells. Thus, the functionality of these tagged proteins is questionable. In particular, the analysis focuses on MDC1 although MDC1 is one of the less functional tagged proteins. For example, MDC1 is reported as less mobile than histones H2B: is this really possible? Could it be an effect of the partial loss of MDC1 functionality? If not, how can such slow mobility be explained? The manuscript also reports that the constitutive interaction between MDC1 and chromatin is mediated by the PST repeat domain of MDC1: again, it is necessary to be careful about this conclusion since the functionality is reduced in the Halo-tagged version.

We appreciate that fully functional fusions may be out of reach, but the limitations need to be acknowledged and discussed. Have alternative tag designs been tried?

We have now included statistical analysis comparing the survival curves for each protein depicted in Figure 2A. The reviewers are correct that ATM, MDC1, 53BP1, and SHLD2 are more sensitive to Zeocin than untagged parental U2OS cells. This indeed indicates that there is some functional impact of HaloTagging on one or more functions of these proteins. While ATM could not be tagged at either terminus, all the other proteins get robustly recruited to DNA damage sites as shown by their foci-forming ability. This largely indicates to us that the impact on function likely relates to events occurring post-recruitment (e.g. affected protein-protein interactions). While this is certainly an important limitation which we have now discussed in the revised manuscript, the types of experiments performed in this manuscript largely only depend on the ability of each protein to get recruited to break sites.

In terms of MDC1 the reviewers are correct that there is impaired function upon

HaloTagging, although these cells respond better to Zeocin treatment than their knockout counterparts. We do make a number of important conclusions from MDC1 and have included some limitations of HaloTag-induced changes in protein function in the Results section. Furthermore, we have included two new pieces of data in Figure S6 demonstrating that HaloTagged wild-type MDC1 can support HR as measured by the DR-GFP assay and that HaloTagged wild-type MDC1 also supports downstream recruitment of 53BP1 to DSB sites. Thus, while the HaloTagged MDC1 does have some sensitivity to Zeocin compared to wild-type cells, it can functionally support DSB repair.

It is important to note that transiently expressed H2B and NLS served as benchmarks for what to expect from a protein that should have a high fraction bound vs. a low fraction bound. However, because of the differences in how the proteins are expressed (transient overexpression vs. endogenous knock-in) it wouldn’t be fair to say much more than that MDC1 and H2B have similar fractions bound. While the fraction of bound MDC1 particles was higher than transiently expressed Halo-H2B (likely due to the transient overexpression of Halo-H2B), bound MDC1 (0.024 m2/s) particles had a higher diffusion rate than bound H2B (0.012 m2/s) which would not be unexpected. Conversely, free MDC1 diffused more slowly than free H2B which would be expected based upon the large difference in protein molecular weight. While we are not the first group to suggest that MDC1, through its PST repeat domain, constitutively interacts with chromatin (PMID: 31729360), we are the first to be able to directly observe constitutively bound MDC1 in living cells, rather than by chromatin-IP.

4) In the section: Functional validation of HaloTagged DDR proteins:

In the absence of zeocin treatment, cells expressing Halo-MDC1 exhibit many spontaneous foci. Is it something known and if not, how can that be explained?

The reviewers are correct in their observation. While not all cells possess these foci without treatment, it is not unusual to observe low levels of foci in nearly all the tagged cell lines and this is not exclusive to Halo-MDC1. These spontaneous foci likely are indicative of low levels of endogenous DNA damage in cells that is bound by MDC1. DSBs occur in every cell of the human body every day, so it is not surprising to observe rapidly dividing cancer cells with small numbers of pre-existing foci in the absence of treatment. It is critically important, however, to point out that in all cases where there are some-pre-existing foci, that foci number is markedly increased upon Zeocin treatment, consistent with the retained ability of these tagged proteins to accumulate at sites of DSBs. A cursory inspection of the literature demonstrates that by immunofluorescence MDC1 can form foci in the absence of exogenous DNA damage (PMID: 18757370; 17158742).

Cells expressing ATM-Halo do not form foci after zeocin treatment: please comment.

Concerning DNA-PK: this protein is highly abundant in the cell; however, it does not form foci after zeocin treatment. Is there an explanation? does it mean that even if the protein is very abundant, very few DNA-PK molecules are present within foci and sufficient for the next steps of NHEJ? Does it form a visible line after micro-irradiation?

We tagged ATM at both termini with HaloTag. Unfortunately, it was not functional when tagged at either terminus. Tagging one terminus interfered with proper localization in the nucleus (Figure 2A). Tagging the other terminus did not interfere with proper localization, however these cells were exquisitely sensitive to Zeocin similar to inhibiting ATM with an ATM kinase inhibitor (Figure S2C & D). This data indicated to us that HaloTagging ATM dramatically impairs function, so for these reasons it is not at all surprising that we do not observe foci in response to Zeocin-induced DSBs.

Concerning DNA-PKcs, we did not observe foci upon Zeocin treatment. This was expected based upon the well-known stoichiometry of DNA-PKcs during NHEJ where only two DNAPKcs proteins are bound at a DSB. Therefore, two molecules would not provide enough fluorescent signal above background to be able to observe this in densely labeled cells. In terms of response to laser-induced breaks, yes, DNA-PKcs does form a visible stripe (Quantified in Figure 4B & Supplemental Figure 4A; Movie provided in Supplemental Movie S1).

5) Page 12, line 206: It is stated that MDC1 has a higher protein abundance than ATM, SHLD1, SHLD2, and SHLD3 but one of the two MDC1 clones analyzed has the lowest protein abundance of all clones analyzed in this study (2400 molecules per cell according to the text and table I). Why are the two MDC1 clones differ so drastically from each other? Confusingly, the big difference between the two clones is not seen in the bar chart in Figure 3B and is not measured by flow cytometry.

Western blots for MDC1 are exceedingly challenging. We attempted to use multiple MDC1 antibodies and multiple transfer conditions. The differences between the two MDC1 clones were a result of somewhat inconsistent transfer onto nitrocellulose membrane. We have now included a statement to this effect in the revised manuscript. The difference in molecules/per cell visible on the bar chart in Figure 3B and the values obtained from flow cytometry experiments are also plotted in the bar chart.

6) Page 12, line 231: It is stated that the adjustment factors for DNA-PKcs are 0.62 and 0.79 but the application of these factors increases the molecule number for clone 1 but decreases it for clone 2 (the latter seems wrong).

We thank the reviewers for pointing out this discrepancy. Indeed, the values after TEV correction for DNA-PKcs C2 were input incorrectly. This has been corrected in the revised manuscript.

7) Single Particle Tracking analysis:

Using analysis of single particle tracking, the authors can measure the diffusive properties of repair proteins. Diffusion is estimated in the presence and in absence of zeocin treatment. Thus, the cells contain many foci: all the traces in the nucleus are analyzed at once, inside and outside foci. The authors then used a 2 population model to fit the distribution of protein displacements.

The power of live-cell single-molecule imaging is that we are able to assess chromatin recruitment of individual DNA repair factors, even if the protein does not accumulate to form a DNA repair focus or prior to DNA repair focus establishment. For this reason, live-cell single-molecule imaging is conceptually very different from characterizing or tracking DNA repair foci and offers a completely new way of analyzing DNA repair by providing a comprehensive picture of how DDR proteins individually behave in both unperturbed conditions and after induction of DNA damage.

The SPT analysis allows the authors to give some interesting mechanistic insight but the authors could extract much more information from their data. Why are the values of Dslow not provided? The authors interpret the slow population as bound to damaged DNA. However, it is known that some proteins diffuse relatively fast inside repair foci, especially if they are able to form liquid-liquid phase separation. In addition, some proteins might diffuse slowly outside of foci, because of non-specific interactions (or even for MDC1 for example). Thus, it is possible that the slow molecules are a mixture of the molecules inside the foci with molecules exhibiting chromatin binding outside of the repair foci.

Thank you for this constructive suggestion. The Dbound data for each individual cell was originally included in the supplementary data, however, we have now moved this into Figure 5. Figure S5D contains the Dbound data compiled from each of three individual replicates.

Indeed, the reviewers are correct that one would potentially expect to observe differences in Dbound between factors that are physically incorporated in nucleosomes (e.g. H2B, Dbound 0.012±0.001 µm2/s), directly interact with nucleosomes or modified histone tails (e.g. MDC1, Dbound 0.024 ± 0.002 µm2/s), or form into liquid-liquid phase separated droplets (e.g. 53BP1, Dbound 0.099 ± 0.012 µm2/s). Our observations are consistent with those expectations, 53BP1 is 8-fold more dynamic in its bound state compared to H2B, which could be a consequence of rapid localized chromatin sampling, or movement of 53BP1 within a phase-separated droplet. These data are also consistent with new residence time analysis data we have included in the revised version of the manuscript where H2B has the residence time, which directly reports on the dissociation rate of the biochemical interaction resulting in the immobilization of the tracked molecule. In contrast, 53BP1 has the lowest residence time suggesting that it interacts with chromatin more dynamically.

There is no visualization of the traces allowing us to see if the slow molecules are indeed inside foci and the fast ones are outside. It would be essential to be able to see this for at least 1 cell for each repair protein.

As mentioned above, the strength of our approach is that we are able to analyze chromatin recruitment of the analyzed DDR factors without a requirement to form DNA repair foci. DNA repair foci likely represent DNA breaks that take particularly long to repair and require extensive processing. Therefore, rather than being limited to repair foci, our approach enables us to monitor individual DDR protein recruitment to DNA damage sites throughout the entire nucleus. To address the dynamics of a protein critical for DNA repair foci formation, we have included a new experiment to analyze 53BP1 at the single-molecule level and simultaneously detecting DNA repair foci. We demonstrate that long lasting static 53BP1 particles can be found inside and not associated with DNA repair foci, which suggests that chromatin association of 53BP1 is not limited to DNA repair foci, which further highlights the strength of our approach. In addition, this experiment demonstrates the 53BP1 molecules can rapidly transition in and out of DNA repair foci as well as transiently become immobile outside of DNA repair foci. These observations demonstrate that 53BP1 has two binding modes in DNA repair foci, likely representing molecules directly associated with chromatin and recruited through phase-separation based interactions, respectively.

Are mixed traces observed, with a slow and a fast part?

Is it possible to estimate the residence time of each protein inside repair foci or on their substrate?

Thank you for the suggestion. As described above we do observe 53BP1 molecules that transition between free and bound states. Reliable quantification of the transition dynamics requires highly accurate particle tracking over extended periods of time and is not feasible with the data generated in this work. A particular issue in eukaryotic cells is the drift of molecules out of the focal plane of the experiment making it challenging to estimate true transition rates. We attempted to analyze the residence time of 53BP1 particles in DNA repair foci using the dual labeling strategy described above. Unfortunately, the number of molecules associated with DNA repair foci is limited making precise quantification beyond the anecdotally observed long binding events challenging. Future studies that are specifically geared towards addressing this question will be carried out.

Are the proteins inside foci exchanging with the rest of the nucleus or are they stuck inside the focus during the entire trace?

Since you use bright JF, it should be possible to have long traces: the authors should show a distribution of the traces' length for each repair protein.

As described above we do observe 53BP1 molecules that transition in an out of DNA repair foci. We analyzed the residence time (equivalent to the track length) for static particles of all proteins which approximates the dissociation rate of the underlying biochemical interaction. We observed an increase in the residence time for 53BP1, SHLD2, and SHLD3 after zeocin treatment suggesting that the biochemical basis of the interaction of these proteins with chromatin is changed by DNA damage induction (for example by chromatin modification or break resection). The observed residence times are limited in length by photobleaching even though we are using bright and photostable JF dyes. Future studies with modified imaging conditions will be necessary to more accurate determine the total binding time of the tagged DNA repair factors.

Do you see a change in protein diffusion in the absence of zeocin treatment and in the presence of zeocin treatment outside of foci?

Because of the nature of live-cell single-molecule imaging, we were not limited to studying DNA repair foci, but were able to use a nucleus-wide approach to monitor how individual DDR proteins. For these experiments, we did not specifically use a marker for DNA repair foci, so this analysis cannot be performed with these particular data sets because there is no reference for where repair foci are located in the nucleus. This type of analysis could indeed be performed in the future but we think this type of analysis is outside the scope of the current manuscript. However, we have included one experiment in Figure S5 where we labeled Halo-53BP1 with JFX650 at the single-molecule level and densely labeled with JF503 to mark 53BP1 foci where demonstrate dynamic associations of 53BP1 inside and outside foci.

The authors also use H2B and NLS as controls. The values obtained should be compared with values found in the literature.

This is a great suggestion. We have included a section in the revised text where we reference values for Halo-H2B and Halo-NLS in the manuscript published in eLife which describes SpotOn analysis by Hansen et al. (PMID: 29300163). For both proteins the fraction of bound particles is comparable: Halo-H2B FBound = 66% vs. ~75% in Hansen et al.; 3XFLAG-Halo-3XNLS FBound = 16% vs. ~10-15% in Hansen et al., but it is important to note that the Halo-3XNLS described in Hansen et al. differs from that used in this manuscript due to the addition of a 3X FLAG tag.

For Halo-H2B we report a DFree of 2.164 ± 0.194 µm2/s vs. ~4 µm2/s in Hansen et al. Similarly, for 3xFLAG-Halo-3xNLS we report a DFree of 3.784 ± 0.512 µm2/s vs. ~10 µm2/sec in Hansen et al. The reported values for the DFree differ between those reported in this manuscript and those reported in Hansen et al. for a couple reasons. Hansen et al. performed stroboscopic photo-activation single particle tracking (spaSPT) in which photoactivatable JF dyes allowed for imaging of ~1 particle per frame on average, while we did not use photoactivatable dyes and had a higher labeling density (~20 ± 5 localizations per frame for Halo-H2B and ~11 ± 1 localizations per frame for Halo-NLS). To determine the free diffusion coefficient of extremely rapidly diffusing molecules like Halo-NLS low labeling density is critical because the tracking algorithm is more likely to connect molecules in close proximity and the likelihood of inadvertently connecting particles incorrectly increases with the rate of diffusion and the labeling density. Importantly, while the increased labeling density in our experiments can impact the absolute value of the free diffusion coefficient of rapidly diffusing molecules, relative diffusion coefficient values are highly reliable. In addition, for the shieldin complex where conclusions were drawn based upon the free diffusion coefficient of the individual subunits labeling density in the nucleus was comparable and very low, which makes these measurements highly reliable.

Finally, it is not clear how the histones H2B are tagged in this study: is it also an endogenous H2B-Halo tag? Is it a stable cell line but not endogenous, or is it a transient transfection?

Both Halo-NLS and Halo-H2B were expressed by transient transfection.

8) Laser micro-irradiation induces massive damage and may not be reflective of physiological encountered DNA damage. Have the authored considered using Cas9-induced DSBs as a defined and targeted DNA damage? I understand that adding such experiments for all proteins would be a massive endeavor, but maybe this could be done for MDC1 and/or the Shieldin complex. Regardless, the limitations of the laser micro-irradiation approach should be discussed.

The application of a versatile tag like HaloTag to study DNA repair using experiments with either sparse or dense protein labeling opens up many exciting opportunities for future research. While we have discussed performing very fast CRISPR on demand, we agree with the reviewers that this would be a massive endeavor and would argue that the addition of these types of experiments are outside the scope of the current manuscript. We have now included a brief discussion in the Results section describing some limitations of using laser micro-irradiation for recruitment analyses.

9) Figure 6 lacks data for the analysis in the presence of zeocin, in the way it was done in the analysis for Figure 5. Such data will corroborate the foci analysis and potentially reveal differences in the recruitment of MDC1 to damaged sites.

We agree with the reviewers and this is an excellent suggestion. We have included single particle tracking of the three different MDC1 mutants in the presence/absence of Zeocin (Figure S6C). The data demonstrates that while the bound fraction of full-length and the BRCT deletion mutant do not change upon Zeocin treatment, there is a significant increase in the bound fraction for the PST deletion mutant, complementing our other observations that MDC1 can interact in at least two ways with chromatin. Furthermore, this new data demonstrates that the PST domain is critical for supporting MDC1’s constitutive association with chromatin, but is not required for MDC1 recruitment to DNA DSBs by its H2AXbinding BRCT domains.

10) What is the evidence that a single molecule can be tracked in Figures 5 and 6, as opposed to a single particle that may be composed of multiple proteins?

For these experiments we label only a small fraction of the overall cellular pool of each protein. It is entirely possible that proteins that are part of larger multi-protein or homomultimers are labeled. These could potentially be distinguished of the individual proteins and larger complexes have distinct diffusion properties. Importantly, the confidence that the vast majority of signals are derived from a single fluorophore is very high. To demonstrate this point, we have included intensity profiles of the analyzed trajectories for MDC1 and RIF1, which are highly static reducing the probability of intersecting tracks leading to additive intensity values.

11) Page 13, line 256: It was surprisingly observed that REV7 and SHLD3 have vastly different recruitment times to laser-induced damage although both factors are known to interact. At the end of the Discussion section, it is speculated that this might reflect REV7's role in TLS but the reader would benefit if such an interpretation was offered earlier in the paper. Moreover, if this interpretation was correct, one would expect that REV7 was recruited to laser damage independently of SHLD3. This should be tested by siRNA-mediated depletion of SHLD3 (or other factors upstream of SHLD3).

These data do not directly report on the absolute order of recruitment of factors to DSBs but rather only report on the relative time to maximal accumulation of each factor to laser-induced DSBs. This particular question becomes much more complicated by issues related to stoichiometry of each factor at breaks. Additionally, because each cell line is normalized to its own peak fluorescence intensity and because laser power was not equally applied for all factors due to dramatic differences in overall protein expression, it would not be appropriate to make a direct comparison between proteins in terms of the absolute number of proteins recruited to laser-induced DSBs. To avoid overinterpreting our results we have removed the more speculative statements regarding the interplay of REV7 and SHLD3 in recruitment to DSBs.

12) Page 14, lines 260-263: It is observed that SHLD1 is not recruited to laser damage although it forms foci after zeocin treatment. The authors then speculate that this might suggest that SHLD1 recruitment is the key regulatory step for a fill-in reaction. This is not clear. Are the authors suggesting that fill-in takes place at zeocin-induced breaks but not at laser-induced breaks? Please clarify.

When we initially performed the laser micro-irradiation experiments for SHLD1, we took images every thirty seconds after irradiation and never could detect SHLD1 recruitment up to ~90 minutes post-irradiation. At the time, we made this speculation to better make sense of this unexpected data. However, we repeated this experiment again using a much longer period of time between images (10 minutes) and successfully detected recruitment of SHLD1 to laser-induced DNA breaks (Figure S4D). This new data suggests our initial imaging conditions were leading to fluorophore bleaching prior to SHLD1 accumulation which impaired our ability to detect it due to low protein abundance in combination with a low amount of protein recruited to break sites. SHLD1 is recruited, but in very low amounts to DSBs which we could visualize, but not easily quantify. This new data is now included in Figure S4D and we have modified the text to account for the new data.

13) Page 14, line 266: The authors state that the measured recruitment kinetics provide insight into the interdependencies of the shieldin complex components. As evidenced by the very early recruitment kinetics of REV7 (possibly due to its role in TLS), this may not be true since the factors could have roles in other processes. Thus, the evaluation of interdependencies requires the measurement of the recruitment kinetics in situations when individual components of the shieldin complex are depleted by siRNA technology. Thus, the authors should assess the recruitment kinetics in cells depleted for SHLD2/3, REV7, or 53BP1/RIF1.

We have changed the wording here in the Results section and in the discussion. This data in its current state can only report on the overall kinetics of Shieldin assembly. We are by no means challenging the well-documented genetic dependencies in Shieldin complex assembly which has been described by several other groups (PMID: 30022168, 29656893, 30022119).

14) Page 17, line 319: SHLD1 is not recruited to chromatin after DSB induction despite its low abundance, and the authors state that this is consistent with its lack of recruitment to laser damage. However, it is clearly shown to form foci at DSB. Moreover, how does this finding fit the author's suggestion from above that SHLD1 is the key regulator of the fill-in reaction?

When we initially performed the laser micro-irradiation experiments for SHLD1, we took images every thirty seconds after irradiation and never could detect recruitment up to ~90 minutes post-irradiation. At the time, we made this speculation to better make sense of this unexpected data. However, we repeated this experiment again using a much longer period of time between images (10 minutes) and successfully detected recruitment of SHLD1 to laser-induced DNA breaks (Figure S4D). This new data suggests our initial imaging conditions were leading to fluorophore bleaching prior to SHLD1 accumulation. SHLD1 is recruited, but in very low amounts to DSBs which we could visualize, but not easily quantify. This new data is now included in the manuscript in Figure S4D and we have modified the text to account for this new data which is consistent with that observed in live-cell imaging of SHLD1 foci.

15) The authors try to address the unexpected findings for SHLD1 (no recruitment to laser damage and no increase in the chromatin-bound fraction after zeocin treatment) in the discussion on page 22, last paragraph, and suggest that SHLD1 may either not bind every DSB where SHLD2 is present or may reside at a DSB in lower copy numbers than SHLD2. Both explanations appear inconsistent with the robust formation of SHLD1 foci at zeocin-induced DSB, the first suggestion can easily be tested with co-localization experiments using HaloTag SHLD1 and SHLD2 ab or vice versa.

We agree with the reviewers that this or a comparable experiment would be critical considering the discrepancy between SHLD1 foci formation after Zeocin and the apparent lack of SHLD1 recruitment to laser micro-irradiated sites. However, the new data presented in Figure S4D where we demonstrate successful detection of SHLD1 recruitment to laser micro-irradiated sites helps to resolve the speculation originally in the text to explain the difference between the two experiments. We have now corrected the text in accordance with the new data.

16) Page 21, line 390: The authors suggest the model that RNF169 delays 53BP1 recruitment to DSBs. Although this interpretation is consistent with the presented data, the model should be tested by 53BP1 recruitment measurements after siRNA-mediated depletion of RNF169.

In the revised version of the manuscript we removed sections where we made bold conclusions regarding the differences between RNF169 and 53BP1 recruitment times to laser micro-irradiated sites. Additional studies are being planned to go further in depth into the interplay between RNF169 and 53BP1 in DSB repair in the future.

[Editors’ note: what follows is the authors’ response to the second round of review.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

[…]

The manuscript is well-written, but some literature reference and information in the methods section are missing.

In conclusion, this is an interesting study that applies novel microscopy techniques to DNA double-strand break repair proteins. However, the study remains somewhat descriptive, which limits the mechanistic insight gained from the study and the analysis remains incomplete in several instances for lack of genetic corroboration of the main conclusion about the Shieldin complex recruitment.

Recommendations for the authors

Essential revisions

1) Line 144: The authors conclude that "the HaloTag does not impact the proper cellular localization of these proteins" based on fluorescence microscopy of the Halo-tagged proteins after JF646 labeling. This conclusion cannot be made, because it would require examination of the localization of the untagged proteins under the same conditions. Please qualify your statement.

In the revised manuscript we have now included an additional statement to qualify this point.

2) Line 146: The authors conclude that "HaloTagging ATM at the N-terminus led to nuclear exclusion". The authors cannot make this conclusion without imaging the untagged ATM under the same condition. Please qualify your statement.

We thank the reviewers for this comment. We HaloTagged ATM at both the C-terminus and N-terminus. While the C-terminally tagged ATM was expressed predominately in the nucleus, the N-terminally tagged ATM was largely excluded from the nucleus. This suggests that tagging one of the termini affects normal ATM subcellular localization. ATM has been long known to be largely localized to the nucleus (Gately et al. Mol Biol Cell, 1998 PMID: 9725899), so it would seem reasonable to conclude that the N-terminal HaloTagging of ATM is what is responsible for its nuclear exclusion even without imaging the untagged protein in living cells. However, we have qualified this statement so as not to state unequivocally that HaloTagging ATM at the N-terminus leads to nuclear exclusion of ATM.

3) Line 183: It cannot be concluded that "most possess full DNA repair functionality" unless the cell lines expressing Halo-tagged proteins are compared to their gene knockout counterparts. This was only performed for 53BP1 and MDC1, where partial functionality was observed. Please qualify your statement.

In the revised version of the manuscript, we have reworded and qualified this statement by stating that these data suggest that most of these proteins retain at least partial DNA repair functionality.

4) Page 13, line 256: It was surprisingly observed that REV7 and SHLD3 have vastly different recruitment times to laser-induced damage although both factors are known to interact. At the end of the Discussion section, it is speculated that this might reflect REV7's role in TLS but the reader would benefit if such an interpretation was offered earlier in the paper. Moreover, if this interpretation was correct, one would expect that REV7 was recruited to laser damage independently of SHLD3. This should have been tested by siRNA-mediated depletion of SHLD3 (or other factors upstream of SHLD3) and this limitation should be explicitly acknowledged in the text.

In the revised manuscript we have added a statement explicitly acknowledging this limitation in the text.

5) Page 14, line 266: The authors state that the measured recruitment kinetics provide insight into the interdependencies of the shieldin complex components. As evidenced by the very early recruitment kinetics of REV7 (possibly due to its role in TLS), this may not be true since the factors could have roles in other processes. Thus, the evaluation of interdependencies requires the measurement of the recruitment kinetics in situation when individual components of the shieldin complex are depleted by siRNA technology. Thus, the authors should assess the recruitment kinetics in cells depleted for SHLD2/3, REV7 or 53BP1/RIF1. The caveat of roles in independent processes should be explicitly mentioned.

In the revised manuscript we have added a statement explicitly acknowledging this limitation in the text as well as describing the caveats of roles of these factors in independent processes.

6) Line 293: It is counter-intuitive that SHLD2 foci depends on SHLD3, which is recruited to foci significantly later than SHLD2. The two-step model suggested in the Discussion to explain this observation is not supported well by the data, because one would expect the initial (pre-SHLD2) step of SHLD3 to also be detected in the LMI experiments. Further, SHLD2 actually dissociates when SHLD3 associates with laser stripes, which is not discussed by the authors. It would strengthen the manuscript if the model could be tested experimentally by the authors.

It is important to note that for these experiments each protein was normalized to its maximal accumulation. To avoid having to normalize the fluorescence intensities, we performed laser micro-irradiation and imaged SHLD2 and SHLD3 using the exact same imaging conditions so we could directly compare between these two proteins and measured absolute fluorescence intensities as a readout of recruitment (Figure S4D). These experiments demonstrate that SHLD3 and SHLD2 are simultaneously recruited to DNA breaks. However, SHLD2 reaches maximal accumulation much earlier than SHLD3 which continues to accumulate in excess of SHLD2 (Figure S4D). In terms of the appearance that SHLD2 dissociates when SHLD3 associates, this appears to be driven by 3-4 cells where there was a reduction in SHLD2 at laser-induced DNA damage sites, while most cells did not have this appearance (Figure S4A and S4D).

7) A similar study was conducted previously (Aleksandrov et al. 2018), which reported recruitment half-times to laser stripes significantly different than the current study for some proteins and in other cases similar half-times. For example, the recruitment times for MDC1, RNF168, RNF169, and 53BP1 was reported by Aleksandrov to be 35s, 78s, 203s, and 307s, where the current study report 77s, 69s, 186s, 669s. It would be in place to reference the previous study and compare findings.

This is an excellent suggestion from the reviewers. We have now included a comparison of our findings to those in Aleksandrov et al. While our results for RNF168 and RNF169 essentially match those in Aleksandrov et al. Our findings for MDC1 and 53BP1 differ ~2-fold. While our experiments used endogenously HaloTagged human 53BP1, Aleksandrov et al. used a HeLa cell line expressing mouse 53BP1, this could provide a reasonable explanation for the differences between the two studies. Additionally, the cell lines we engineered exclusively express HaloTagged proteins while the cell lines used in Aleksandrov et al. express eGFP-tagged proteins in the presence of the wild-type unedited proteins. Therefore, the differences in the recruitment of MDC1 and 53BP1 between the two studies could be a consequence of differences in the protein sequence, gene dosage or competition with the endogenous protein leading to altered recruitment in Aleksandrov et al. compared to what we observe in the present study. We discuss these differences in the revised manuscript.

8) It is surprising that Halo-H2B only displays 66% chromatin binding. The assumption is that JF646 binds irreversibly to the Halo-tag, but is it possible that some free JF646 is present and gives rise to the "free" pool of fluorophores? This could easily be tested by formaldehyde fixation which should give 100% chromatin binding if no free JF646 is present.

It is important to note that Halo-H2B was transiently overexpressed in the presence of wildtype H2B, so it is not unreasonable to assume that the cellular pool of H2B is higher leading to a competition of H2B and Halo-H2B competing for inclusion into a limited number of nucleosomes and therefore the values reported here may artificially lower than expected. Despite this, the value we obtained for Halo-H2B in terms of the fraction of bound particles (~66%) is comparable to that described in Hansen et al. (Hansen et al., eLife. 2018) where they observed an Fbound of ~75% using Spot-On analysis and transiently expressing Halo-H2B using the same plasmid. The purpose of using the Halo-H2B was only to have a control for what to expect to observe if a protein is largely chromatin bound as opposed to the 3x-FLAGHalo-NLS which was used as a control for the lower bound of what one would expect if a protein was largely freely diffusing. We do not believe free fluorophores contribute to our analysis, since we do not observe nuclear particles in control cells that do not express a Halotagged protein.

9) Line 331: A two-state diffusion model is assumed where particles either freely diffuse or are chromatin bound. How would the conclusions be affected if a third state was allowed where a protein is part of a slow diffusing macromolecular complex.

Inclusion of a third state typically has a minimal effect on the static fraction of proteins we have studied in the past. The third state typically has intermediate mobility and comes at the expense of the fraction of freely diffusing molecules. In our past work on telomerase, we have explicitly used a three-state model because the TERT protein can be either freely diffusing or part of the large telomerase RNP, which significantly reduces its mobility (Klump et al. Cell Reports 2023). Similarly, we have used a three state model to analyze 53BP1 trajectories inside and outside of DNA repair foci in this updated manuscript, assuming that 53BP1 can be either freely diffusing, directly bound to chromatin, or part of phase-separated DNA repair foci.

10) Figure 5: The scatter plots should be displayed as "super plots" where data points for each of the 3-4 independent experiments are presented in different colors/symbols (Lord et al. 2020). This would reveal any systematic differences between experiments. For example, it looks like RNF169 data points can be divided into two populations.

We have now plotted these plots as super plots as the reviewers have suggested in the revised version of the manuscript.

11) Page 21, line 390: The authors suggest the model that RNF169 delays 53BP1 recruitment to DSBs. Although this interpretation is consistent with the presented data, the model could be tested by 53BP1 recruitment measurements after siRNA-mediated depletion of RNF169. This limitation should be explicitly acknowledged in the text.

While we had suggested in the original submission that RNF169 delays 53BP1 recruitment, this language was removed in the second submission. It is not clear to us, which statement the reviewer is referring to.

12) The sources of several plasmids are missing e.g. pX330 in line 665.

The sources for px330, pFastBac Dual, and pRK2 have now been included in the Material and Methods section.

13) A table with oligonucleotides and gRNAs used in the study should be included.

A table with oligonucleotides and gRNAs used in this study has been included in the resubmission of this manuscript.

14) Line 511: Which data allow the authors to conclude that the PST domain of MDC1 "facilitates DDR signal amplification"?

We admit that this may have been strongly worded. Therefore, we have removed the term “signal amplification” and kept the statement that it is important for DSB repair based upon results from the HR DR-GFP assay.

15) The authors conclude that MDC1 and RIF1 are constitutively associated with chromatin. If this is the case then one might expect the localization of MDC1 and RIF1 to follow that of condensed chromosome when cells progress from interphase into mitosis. Indeed, there is evidence for this for RIF1 (Watts et al. 2020), but for MDC1 I could not find such evidence in the literature. However, I suspect that the authors in their data have images of mitotic cells that could answer this question.

We thank the reviewers for this thoughtful comment. This is actually an exciting area of ongoing inquiry that is nearing completion and submission for publication. Because of this we feel that presenting data related to this question is outside of the scope of the current manuscript.

16) To quantify the mobility inside foci versus outside foci, and the flux of proteins in and out of foci, would it be possible to make a density map of all the repair proteins in the nucleus? Using this density map, the foci appear clearly, and it is then possible to separate trajectories within foci from those outside foci.

We have added a new analysis that filters trajectories using a mask of the nuclear 53BP1 foci. Trajectories are separated into two groups, one in which trajectories overlap with DNA repair foci and a second group of tracks that never co-localizes with repair foci and analyzed the data using a three stated model assuming 53BP1 can be freely diffusing, statically bound to chromatin, or part of phase separated DNA repair foci (which would move with an intermediate diffusion coefficient). In DNA repair foci 53BP1 particles moving with an intermediate diffusion coefficient were enriched compared to particles that did not colocalize with DNA repair foci, consistent with the hypothesis that 53BP1 can be recruited to DNA repair foci by a phase separation mechanism.

17) Figure S5B: After zeocin treatment: is it possible that the lines between foci are misslinking?

In the context of single particle tracking miss-linking can occur, for instance if a static particle is failed to be detected in a given frame due to signal fluctuations (for example as a result of photo-blinking) a trajectory can be incorrectly linked to a nearby molecule. In general, the number of incorrect steps included in our analysis is minimal compared to correct linkages and therefor do not impact global step distance analysis like we do with the Spot-On tool. Correct linkage is extremely important for residence time analysis because long binding events can be chopped up into short trajectories. For this reasons, we averaged multiple frames together to amplify static binding events and minimize the contribution of signal fluctuations.

https://doi.org/10.7554/eLife.87086.sa2

Article and author information

Author details

  1. Joshua R Heyza

    Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, United States
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Writing - original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0847-0501
  2. Mariia Mikhova

    1. Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, United States
    2. Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, United States
    Contribution
    Formal analysis, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Aastha Bahl

    Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  4. David G Broadbent

    1. Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, United States
    2. College of Osteopathic Medicine, Michigan State University, East Lansing, United States
    3. Department of Physiology, Michigan State University, East Lansing, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0940-1068
  5. Jens C Schmidt

    1. Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, United States
    2. Department of Obstetrics, Gynecology, and Reproductive Biology, Michigan State University, East Lansing, United States
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Methodology, Project administration, Writing – review and editing
    For correspondence
    schmi706@msu.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9061-7853

Funding

National Institutes of Health (F32GM139292)

  • Joshua R Heyza

National Institutes of Health (DP2GM142307)

  • Jens C Schmidt

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are grateful to Dr. Kathy Meek for providing the DNA-PKcs antibody. We thank Dr. Daniel T Youmans and Dr. Thomas R Cech for providing the plasmid for recombinant production of the 3xFLAG-HaloTag protein. We thank Dr. Eric Patrick for contributing to the purification of the 3xFLAG-HaloTag protein. Funding This work was funded, by NIH grants F32GM139292 to JRH and DP2GM142307 to JCS. This work was supported by the microscopy and flow cytometry cores in the Institute of Quantitative Health Sciences and Engineering at Michigan State University. The MSU Flow Cytometry Core facility is funded, in part, through the financial support of Michigan State University’s Office of Research & Innovation, College of Osteopathic Medicine, and College of Human Medicine.

Senior Editor

  1. Detlef Weigel, Max Planck Institute for Biology Tübingen, Germany

Reviewing Editor

  1. Wolf-Dietrich Heyer, University of California, Davis, United States

Reviewer

  1. Judith Miné-Hattab, Institut Curie, PSL University, Sorbonne Université, CNRS, Nuclear Dynamics, France

Version history

  1. Preprint posted: June 9, 2022 (view preprint)
  2. Received: February 18, 2023
  3. Accepted: June 20, 2023
  4. Accepted Manuscript published: June 21, 2023 (version 1)
  5. Version of Record published: July 4, 2023 (version 2)

Copyright

© 2023, Heyza et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,011
    Page views
  • 60
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Joshua R Heyza
  2. Mariia Mikhova
  3. Aastha Bahl
  4. David G Broadbent
  5. Jens C Schmidt
(2023)
Systematic analysis of the molecular and biophysical properties of key DNA damage response factors
eLife 12:e87086.
https://doi.org/10.7554/eLife.87086

Further reading

    1. Cell Biology
    2. Chromosomes and Gene Expression
    Maikel Castellano-Pozo, Georgios Sioutas ... Enrique Martinez-Perez
    Short Report Updated

    The cohesin complex plays essential roles in chromosome segregation, 3D genome organisation, and DNA damage repair through its ability to modify DNA topology. In higher eukaryotes, meiotic chromosome function, and therefore fertility, requires cohesin complexes containing meiosis-specific kleisin subunits: REC8 and RAD21L in mammals and REC-8 and COH-3/4 in Caenorhabditis elegans. How these complexes perform the multiple functions of cohesin during meiosis and whether this involves different modes of DNA binding or dynamic association with chromosomes is poorly understood. Combining time-resolved methods of protein removal with live imaging and exploiting the temporospatial organisation of the C. elegans germline, we show that REC-8 complexes provide sister chromatid cohesion (SCC) and DNA repair, while COH-3/4 complexes control higher-order chromosome structure. High-abundance COH-3/4 complexes associate dynamically with individual chromatids in a manner dependent on cohesin loading (SCC-2) and removal (WAPL-1) factors. In contrast, low-abundance REC-8 complexes associate stably with chromosomes, tethering sister chromatids from S-phase until the meiotic divisions. Our results reveal that kleisin identity determines the function of meiotic cohesin by controlling the mode and regulation of cohesin–DNA association, and are consistent with a model in which SCC and DNA looping are performed by variant cohesin complexes that coexist on chromosomes.

    1. Chromosomes and Gene Expression
    2. Developmental Biology
    Airat Ibragimov, Xin Yang Bing ... Paul Schedl
    Research Article Updated

    Though long non-coding RNAs (lncRNAs) represent a substantial fraction of the Pol II transcripts in multicellular animals, only a few have known functions. Here we report that the blocking activity of the Bithorax complex (BX-C) Fub-1 boundary is segmentally regulated by its own lncRNA. The Fub-1 boundary is located between the Ultrabithorax (Ubx) gene and the bxd/pbx regulatory domain, which is responsible for regulating Ubx expression in parasegment PS6/segment A1. Fub-1 consists of two hypersensitive sites, HS1 and HS2. HS1 is an insulator while HS2 functions primarily as an lncRNA promoter. To activate Ubx expression in PS6/A1, enhancers in the bxd/pbx domain must be able to bypass Fub-1 blocking activity. We show that the expression of the Fub-1 lncRNAs in PS6/A1 from the HS2 promoter inactivates Fub-1 insulating activity. Inactivation is due to read-through as the HS2 promoter must be directed toward HS1 to disrupt blocking.