Ctp1 protein–DNA filaments promote DNA bridging and DNA double-strand break repair

The Ctp1 protein in Schizosaccharomyces pombe is essential for DNA double-strand break (DSB) repair by homologous recombination. Fission yeast Ctp1 and its budding yeast (Sae2) and human (CtIP) homologs control Mre11–Rad50–Nbs1 nuclease complex activity and harbor DNA-binding and -bridging activities. However, the molecular basis for Ctp1–DNA transactions remains undefined. Here, we report atomic force microscopy (AFM) imaging of S. pombe Ctp1–DNA complexes revealing that Ctp1 polymerizes on dsDNA molecules and forms synaptic filaments that bridge two dsDNA strands. We observed that Ctp1 DNA filaments are typified by an average filament length of ∼180 bp of dsDNA and a Ctp1 tetramer footprint of ∼15 bp. Biochemical results characterizing Ctp1 variants with impaired DNA-binding or -bridging properties were consistent with Ctp1-mediated DNA bridging requiring the intact and correctly folded Ctp1 tetramer. Furthermore, mutations altering Ctp1 oligomerization and DNA bridging in vitro conferred cell sensitivity to DSB-producing agents. Together, these results support an important role for Ctp1-regulated DNA strand coordination required for DNA DSB repair in S. pombe.

Although it remains unclear how DNA repair intermediates are coordinated during adduct removal and resection, the DNA-binding properties of Ctp1 suggest that it could play a role in DNA strand coordination (18 -20). Specifically, S. pombe Ctp1 and its budding yeast counterpart Sae2 are oligomeric proteins that bind DNA and can bridge two dsDNA strands together (18,21). As such, Ctp1 could protect DNA ends from nucleolytic degradation (22), facilitate association of the DSB with the sister chromatid for homologous repair (23), and/or ensure that coincident resection of DSBs occurs, a role ascribed to S. cerevisiae Sae2 (24). DNA-binding sites critical for DSB repair in S. pombe reside at the extreme termini of Ctp1. The conserved N terminus of Ctp1 forms a homotetramer composed of a helical domain that oligomerizes into a ϳ120-Å-long parallel four-helical bundle called the tetrameric helical dimer of dimers (THDD), with DNA-binding sites located in the coiled-coil regions (18,25). C-terminal to the THDD, Ctp1 contains a low complexity intrinsically disordered C-terminal region containing phosphorylated SXT motifs that tether Ctp1 to MRN through Nbs1 (4). Additional functionally critical DNA-binding interaction sites are found at the extreme C terminus in the conserved Sae2-like domain, which harbors a conserved RHR DNA-binding motif (18). In addition, alleles of the human CtIP homolog lacking the C-terminal DNA-binding domain are dominant-negative for Jawad-Seckel syndrome (26), suggesting that the DNA-binding properties of Ctp1 Sae2/CtIP are important for maintaining genome stability. Despite the apparent importance of DNA binding by Ctp1, the molecular basis for Ctp1 DNA-binding and bridging activities remains undefined.
Here, we combine atomic force microscopy (AFM) with biochemical and in vivo studies to characterize WT and Ctp1 mutants that have impaired DNA-binding and/or -bridging properties. We report that Ctp1 polymerizes on dsDNA molecules, forming synaptic filaments that bridge two dsDNA strands, with extended filament lengths spanning ϳ180 bp of DNA and an average Ctp1 tetramer footprint of ϳ15 bp. Ctp1mediated DNA bridging in vitro requires not only the previously reported DNA-binding motifs at N and C termini (18), but also the intact Ctp1 tetramer. Mutations altering Ctp1 oligomerization and DNA bridging in vitro also undermine Ctp1 DSB repair functions in S. pombe. Altogether, these results underline new functional roles for Ctp1 protein-DNA bridging filaments in the coordination of DNA repair intermediates during homologous recombination DSB repair, and they provide a mechanistic basis for the DNA repair defect that causes Jawad-Seckel syndrome in humans.

Molecular basis for Ctp1 DNA binding and bridging
We employed AFM to capture images of unliganded Ctp1 and Ctp1-DNA complexes (Fig. 1, A and B). In the absence of DNA, WT Ctp1 molecules appear as short, diffuse peaks, coupled to regions of lower height, consistent with a predicted intrinsically disordered C-terminal region (Fig. S1A) (18,27). The volume of proteins in AFM images depend linearly on their molecular weight (28 -30); however, the low height of disordered regions of Ctp1 makes it difficult to determine an accurate volume of the proteins. Consequently, we cross-linked Ctp1 with glutaraldehyde prior to deposition, and the crosslinked Ctp1 exhibits taller, defined peaks (Fig. S1B). Based on a standard curve of AFM volume versus molecular weight, this volume corresponds to a molecular weight of 124 kDa (4 Ctp1 molecules) (28 -30), which is consistent with Ctp1 (M r ϳ132 kDa) being predominantly tetrameric (Fig. S1C), consistent with multi-angle light scattering experiments (18).
To define global properties of Ctp1-DNA interactions, we imaged Ctp1 bound to relaxed plasmid DNA. Given the

Ctp1 protein-DNA filaments promote repair
decrease in background noise caused by the intrinsically disordered region of Ctp1 in the presence of glutaraldehyde, we attempted to image Ctp1-DNA complexes in the presence of glutaraldehyde; however, we did not observe any complexes. Further analysis of the Ctp1-DNA interaction by electrophoretic mobility shift assays (EMSAs) demonstrated that glutaraldehyde impaired the interaction of Ctp1 with DNA (data not shown). As a result, we conducted AFM imaging of the Ctp1-DNA interactions without chemical cross-linkers. In the absence of Ctp1, the DNA appears as an open circular structure with no high features, typical of relaxed plasmid DNA (31,32). In contrast, in the presence of Ctp1, we observe tracts of protein bridging two dsDNA molecules together (DNA bridging; 60% Ctp1-DNA complexes) as well as tracts binding along a single dsDNA (DNA binding) (Fig. 1, A and B), similar to that seen in AFM images of H-NS-DNA and MutL-␣-DNA complexes (31,32). Most bridging events (75%) are intramolecular ("zippered" plasmids), but intermolecular plasmid-to-plasmid bridging events are also observed (Fig. S2A).
Analysis of the distribution of Ctp1 polymer lengths on DNA reveals that the average length of the Ctp1 protein-DNA filament tracts for the DNA-bridging events is approximately twice as long as that of the DNA-binding events ( Fig. 2A). This result indicates that DNA bridging by Ctp1 promotes its cooperative assembly to "zipper up" two dsDNA chains, resulting in formation of elongated protein-DNA filaments. Notably, although the tracts along one dsDNA show no regular repeating features, the Ctp1-DNA bridging filaments show recurrent striations (Fig. 2, B and C). Quantitative analysis of the interpeak distances within these DNA-bridging complexes show quantized distributions of ϳ10, 15, 20, and 30 nm (Fig. 2D). Given that AFM volume analysis shows that free Ctp1 exists primarily as a tetramer (Fig. S1C), these results suggest that the 5-nm periodicity represents individual tetramers of Ctp1. Interestingly, the volume/length ratio for Ctp1-DNA tracts on one dsDNA is identical to that for bridging tracts (Fig. S2B), suggesting that Ctp1 also has a 5-nm footprint when binding one dsDNA molecule. Furthermore, these results suggest that Ctp1 binding to one dsDNA may provide a nucleation site for bridging. Consistent with this idea, we also observe a greater percentage of Ctp1-bridging events relative to binding to one dsDNA throughout the AFM images (

Ctp1 protein-DNA filaments promote repair Ctp1 oligomerization and N-terminal domain folding facilitates DNA bridging
Based on our previous work, DNA binding and bridging requires conserved residues both in the THDD domain and C-terminal Sae2-like domain bearing an RHR motif of Ctp1 (18) (Fig. 3A); however, the precise roles of Ctp1 oligomeriza-tion in DNA-binding and -bridging activities are unclear. Therefore, to probe the roles of Ctp1 oligomerization on Ctp1 DNA binding and function, we engineered a series of THDD variants that alter Ctp1 oligomerization. Two N-terminal deletions convert the native Ctp1 tetramer into dimeric (Ctp1-N⌬20) or monomeric forms (Ctp1-N⌬60), as assessed by ana-

Ctp1 protein-DNA filaments promote repair
lytical gel-filtration chromatography (Fig. S3, A and B). Based on the crystal structure of the Ctp1 helical bundle (18), we further engineered a disulfide bond to replace an "a" heptad position of the Ctp1 leucine zipper repeat (Ctp1 variant L51C). This strategy covalently cross-links full-length tetrameric (L51C) and dimeric proteins (N⌬20-L51C) with a stable disulfide (Fig.  3A), visible as dimeric forms on an SDS-polyacrylamide gel even under denaturing and reducing conditions (Fig. S3A). An additional THDD variant, containing an N-terminal maltosebinding protein (MBP) fusion with a fixed-arm linker that does not inhibit tetramerization (18) but potentially blocks DNA access to the THDD, was used to assess the overall necessity of the THDD for Ctp1-DNA interactions. All tetrameric and dimeric forms of Ctp1 retain both the N-terminal and C-terminal DNA-binding motifs.
compared with WT Ctp1, indicating that the leucine zipper conformation is important for native DNA binding. Furthermore, the fusion of the bulky MBP protein to the Ctp1 N terminus (N-MBP) causes a ϳ5-fold decrease in affinity of Ctp1 on duplex DNA (K d ϭ 730 Ϯ 44 nM, h ϭ 1.8 Ϯ 0.2) (Fig. 3D). As this MBP fusion construct is tetrameric (18), the globular MBP protein may sterically block the THDD DNA-binding site rather than impair oligomerization. Taken together, these results demonstrate that access to the Ctp1 THDD and THDD oligomerization are important for stable DNA binding, whereas covalent cross-links that restrict the conformational flexibility of the Ctp1 N-terminal coiled-coil region modulate DNA interactions.
In our EMSA binding experiments, we reproducibly observe progressive supershifting of DNA-protein complexes with increasing addition of Ctp1, consistent with multiple Ctp1 proteins binding or bridging DNA and/or multiple DNA-binding sites in the Ctp1 tetramer mediating tethering of DNA molecules (18) (Fig. 3A). To examine the basis for multisite DNA binding, we evaluated the influence of Ctp1 oligomerization on its DNA-bridging activity with an in vitro DNA affinity pulldown assay (Fig. 4A) (18). Robust DNA bridging was observed for full-length Ctp1, but DNA-bridging activity was severely impaired for the N⌬20-L51C, N⌬60, and N-MBP variants. Unexpectedly, we found that although the N⌬20 construct contains both N-and C-terminal DNA-binding sites and thus retains residual DNA-bridging activity, the disulfide-bonded form (N⌬20-L51C) of this deletion that displays moderate DNA-binding activity (Fig. 4, B and C) has no detectable bridging activity (Fig. 4, C and D).
AFM imaging of mutant Ctp1 proteins in the presence of DNA was consistent with DNA-binding and -bridging in vitro assays (Fig. S4, A and B). N⌬20-L51C binds circular plasmid DNA (Fig. S4A) but did not display filamentous bridging archi-tectures typical of the WT protein. By comparison, N⌬60 did not appreciably interact with plasmid DNA under the conditions of AFM imaging (Fig. S4B), consistent with the decreased DNA-binding affinity of this mutant. Analysis of peak volumes in the AFM images indicates that Ctp1 N⌬20-L51C and N⌬60 have an approximate molecular weight of 78 and 34 kDa, consistent with primarily dimeric (N⌬20-L51C, calculated molecular weight ϭ 61 kDa) and monomeric species (N⌬60, calculated molecular weight ϭ 26 kDa), under AFM experimental Ctp1 protein-DNA filaments promote repair conditions (Fig. S4, C and D). Together, these results indicate that both oligomerization and mobility of the Ctp1 N-terminal coiled-coiled domain are critical for efficient DNA bridging. Moreover, the N⌬20-L51C mutant has unique characteristics in that it retains moderate DNA binding (Fig. 3D) but ablates DNA bridging (Fig. 4C) in vitro.

DNA-bridging mutants compromise Ctp1 genome protective functions
To probe the biological importance of DNA binding and bridging in DSB repair, we examined the effects of Ctp1 DNAbinding and -bridging mutations on clastogen sensitivity of S. pombe. Immunoblotting of the C-terminal 5x-FLAG epitope tag in ctp1 ϩ strains showed that mutant proteins were present at similar levels to WT (Fig. S5). The ctp1 ϩ L51C and ctp1 ϩ N⌬20 display growth phenotypes after DNA damage comparable with WT. In contrast, the N⌬20-L51C mutation exhibits acute sensitivity to all DNA-damaging agents evaluated (Fig. 5 and Fig. S5). Given that the N⌬20-L51C mutant binds DNA (Fig. 3D) but does not efficiently bridge DNA (Fig. 4C), these data suggest that the ability of Ctp1 to form oligomeric tracts that bridge two dsDNAs together plays an important role in DSB repair.

Discussion
Ctp1 is a critical component of the DNA DSB repair machinery, engaging with the MRN complex to initiate resection of the DNA DSB (4 -6); however, the mechanism of Ctp1 action in this role is poorly defined. For efficient repair to occur, Ctp1 must form oligomers capable of efficient DNA-binding and -bridging activity (Fig. 6) (18). Based on our AFM data (Figs. 1 and 2), this bridging action occurs through the formation of protein-DNA filaments. At the gross structural level defined by AFM, the Ctp1 filaments reported herein are reminiscent of bacterial nucleoid-associated protein H-NS and the mismatch repair protein Mlh1-Pms1 from yeast, which form bridging filaments (31,32). Protein-DNA filaments are an architectural feature frequently found to regulate genomic structure and repair (1,2). In DSB repair, HR enlists multiple RPA and Rad51 molecules to bind ssDNA and dsDNA, creating presynaptic filaments in preparation for homologous pairing (33). In nonhomologous end-joining, the XRCC4-XLF filament bridges two DNA molecules independent of DNA ends to maintain DNA DSBs in proximity for repair (34). Ctp1 may play an analogous role in HR and microhomology-mediated end-joining (MMEJ) repair, because in humans, HR and MMEJ employ CtIP tetramers and dimers, respectively (25). Our studies show that both dimeric and tetrameric Ctp1 are capable of DNA bridging (Fig.  4C) and that mutations altering DNA bridging exhibit marked DNA damage sensitivity (Fig. 5), implicating DNA bridging as an important function of Ctp1.
Ctp1-DNA fibers typically extended to much longer tracts of regularly repeated peaks when Ctp1 interacted with two dsDNA molecules in a bridging event, as compared with a binding event (Fig. 2), suggesting a cooperative mechanism of binding. The shortest repeat of bridging filaments observed by AFM is 10 nm or ϳ30 bp (10-nm length of DNA at 34 Å/10 bp for B-DNA). The 5-nm Ctp1-DNA footprint indicates that DNAbridging events involve on average ϳ12 Ctp1 tetramers, whereas DNA-binding events involve ϳ6 Ctp1 tetramers (Fig.  2, A and D). In addition, the overall DNA length is unchanged by bridging (Fig. S2, C and D). Because the maximum particle dimension of free Ctp1 based on small-angle X-ray scattering analysis is 26 nm (Fig. S6A) (18), these observations imply that Ctp1 undergoes a conformational compaction during DNA binding. Structural compaction likely takes place within the flexible, intrinsically disordered C terminus of Ctp1, which   When a DSB occurs, Ctp1 associates with MRN at the break site and may initiate nucleation. Ctp1 filaments can then form and may bridge across the DSB (intramolecular bridging), adjacent to the break site or may bridge between damaged DNA and the sister chromatid (intermolecular bridging), which would stabilize the repair complex and stimulate resection by Mre11. Ctp1 filaments may also bridge across sister chromatids to stabilize recombination intermediates.

Ctp1 protein-DNA filaments promote repair
could take on a variety of conformations when interacting with DNA (Fig. S6B). This change in structure of the Ctp1 C-terminal tails might facilitate a spatial search during DNA bridging. The Ctp1 N-terminal coiled-coil domain may also have important structural rearrangements upon DNA bridging, as locking the Ctp1 coiled-coil in place with a disulfide bond (L51C mutation) abolishes DNA bridging in Ctp1 N⌬20 (Fig. 4C). A critical role for the conserved C-terminal RHR motifs in the Sae2-like domain in mediating DNA binding (18,25) suggests that this region is likely also important for filament formation. Understanding the precise nature of Ctp1-DNA interactions will require more detailed high-resolution structural analysis of Ctp1-DNA complexes by X-ray crystallography or cryo-EM.
Biologically, during coordination of the DNA repair intermediates in both HR and MMEJ, the short tracts of Ctp1 along a single dsDNA molecule may serve to nucleate the extended tracts of Ctp1 filaments through cooperative DNA binding (Fig.  6). Nucleation of Ctp1-DNA filaments may be controlled by MRN complex binding at the DSB (4). Ctp1 does not require free DNA ends for bridging (18), and DNA bridging occurs with relaxed circular plasmid DNA (Fig. 1). Ctp1 can furthermore mediate both intra-and intermolecular bridging (Fig. 1B and  Fig. S2A). These observations provide possible explanations for results showing that the Ctp1 homolog Sae2 is required for in vivo intrachromosomal DNA bridging of HO-induced DSBs (21). Ctp1 filaments may bridge across sister chromatids to coordinate homologous templates or within recombination intermediates in HR (Fig. 6). Tethering DNA repair intermediates adjacent to the break site by Ctp1 filaments could protect DNA from unregulated nucleolytic degradation while also regulating MRN activity near the DSB ends to facilitate coincident (coordinated) DSB resection, as observed for Sae2 (24). The tethering of DNA by Ctp1 may also be important for restricting mobility of broken DNA ends during repair, effectively stitching DNA ends while repair proteins heal the break, and to possibly regulate subsequent strand invasion steps. That the MRN complex can also bridge DNA ends (7,(35)(36)(37) indicates that there are distinct functional requirements to coordinate DSB intermediates before, during, and after DNA end processing and that multiple strand coordination activities control the initiation of HR.
The precise roles of Ctp1-DNA filament architectures in regulating the removal of protein-blocked DNA termini by MRN remain to be determined. We posit that Ctp1 protein-DNA filament formation may limit an otherwise distributive MRN endonuclease activity to cut only near protein-blocked termini where filaments cannot extend. Ctp1, Sae2, and CtIP also exhibit structure-specific DNA end-binding properties (10,12,18,19), so the specific MRN interaction with Ctp1/ Sae2/CtIP on DNA ends may alternatively dictate site-specific cleavage near protein-blocked DNA ends. Future structural work investigating the complex interplay between Ctp1, DNA, and the MRN complex will be required to more clearly define the role of Ctp1-bridging filaments in regulating MRN activity.

Protein expression and purification
Ctp1 WT and N⌬60 were expressed and purified as described previously (18) with a final purification on a 16/60 S200 size-exclusion chromatography column (GE Healthcare) in 20 mM Tris, pH 7.5, 500 mM NaCl, and 0.1% ␤-mercaptoethanol. Ctp1 N⌬20 was expressed and purified as a His 6 -tagged fusion from pMCSG7 (39) identical to Ctp1 N⌬60. L51C point mutations were introduced with QuikChange mutagenesis (Agilent Technologies, Santa Clara, CA) into both Ctp1 WT and N⌬20. Point mutations were expressed and purified identical to their WT counterparts. MBP-Ctp1 was expressed and purified as described previously for MBP-Ctp1 fusions generated in the pMALX vector (18,38).

Atomic force microscopy
Protein-DNA complexes were made by incubating 800 nM Ctp1 with 1 nM pHOT1-relaxed plasmid (TopoGEN Inc., Buena Vista, CA) for 30 min at room temperature in binding buffer (40 mM Hepes, pH 7, 5 mM MgCl 2 , 25 mM NaCl, 0.1 mM tris(2-carboxyethyl)phosphine) in a total volume of 20 l. The entire reaction was deposited on freshly cleaved mica (Ted Pella, Inc., Redding, CA) for less than 1 min and then rinsed with 800 l of molecular biology grade water. The mica surface was then blotted dry and fully dried by a nitrogen stream. Images were captured in air with a Nanoscope IIIa (Digital Instruments, Santa Barbara, CA) microscope in tapping mode using PointProbePlus-NCL silicon probes (Nanosensors, Neuchatel, Switzerland). Images were collected at a size of 2 ϫ 2 m.

Image processing and analysis
Image Metrics version 1.13 (in-house designed MatLab (MathWorks Inc., Natick, MA) program) was used for image processing and analysis. Standard image processing utilized plane subtraction and flattening. Analysis included measuring DNA contour lengths and protein-DNA complex lengths, heights, and volumes. Bridging events were defined as protein interacting with two molecules of dsDNA, whereas binding events were defined as protein interacting with one molecule of dsDNA. Particle analysis to determine bridging versus binding events included protein-DNA complexes with a maximum height of Ն0.6 nm, as this is greater than 3 times the average DNA height of 0.20 Ϯ 0.0008 nm determined by image analysis of DNA particles only; n values are indicated in the figure legends. Additional analysis for Fig. S2B used images that had defined interpeak distances as shown in Fig. 2, B and C. Statistical analyses were carried out using GraphPad Prism version 7.0 (GraphPad Software Inc., La Jolla, CA). Molecular weights of Ctp1-DNA complexes were derived from volumes as determined by ImageMetrics, using a published standard curve (28).

DNA-bridging assays
DNA-bridging assays were carried out as described previously (18), with the following changes. The 500-bp dsDNA substrate used was that described in the DNA-binding assays. 10 l of Streptavidin Sepharose High Performance resin (GE Healthcare) was used in place of magnetic beads, and beads were collected by 2-min spins at 1000 ϫ g in a microcentrifuge. Reactions were resolved on 4 -20% Novex TBE gels (Invitrogen). Experiments (n ϭ 3 for each Ctp1 protein) were imaged and analyzed as described for the DNA-binding assay. Total intensity was determined as intensity in reactions of Ctp1 ϩ DNA minus the intensity in DNA-only reactions. Statistical analyses (t test, two-tailed) were performed by GraphPad Prism version 7.0a (GraphPad Software Inc.).

Strain construction
Methods and growth media for S. pombe genetics followed standard protocols (39). All strains used in this study are listed in Table S1. Point mutations were inserted by QuikChange mutagenesis (Agilent Technologies). Strains were constructed as described previously (18), and sequencing verified the presence of correct truncations and point mutations.

Clastogen sensitivity assays
Assays were performed as described previously (18) and treated with clastogens as labeled in Fig. 6 and Fig. S4.