DNA end configurations dictate synaptic complex formation during NHEJ-mediated end bridging

Abstract

Synapsis, the physical bridging of two broken DNA ends, is a critical step in non-homologous DNA end joining (NHEJ), the primary pathway for repairing DNA double-strand breaks (DSBs) in vertebrates. While NHEJ synapsis has been well characterized using blunt-ended DNA, how diverse DNA end configurations influence synaptic complex formation remains much less clear. Here, using single-molecule FRET, we show that end configurations play a decisive role in synapsis pathway choice, with end pairing compensating for XLF in facilitating synaptic complex transitions. Notably, 3-nucleotide (nt) microhomology at DNA ends significantly enhances close synaptic (CS) complex formation mediated by Ku70-Ku80 (Ku) and XRCC4-Ligase IV (X4L4), even in the absence of XLF. Although XLF is nonessential for synapsis with 3-nt microhomology, it further promotes CS formation when present. The FRET analysis reveals that the single-stranded and double-stranded junctions of the DNA ends are in close contact within the primary CS complex, while the overhangs flip out of the duplex, potentially stabilizing the complex. These findings underscore the pivotal role of DNA end configurations in regulating synapsis and their broader implications for NHEJ repair efficiency and fidelity.

DNA double-strand breaks are primarily repaired by non-homologous end joining, which requires synapsis of broken DNA ends. Here, the authors show that DNA end configurations, particularly short microhomologies, dictate synaptic complex formation and can bypass XLF to promote close end joining.

Introduction

DNA double-strand break (DSB) is the most severe DNA damage in cells. DSB can be caused by both endogenous factors, such as active stresses from common metabolic processes, and exogenous reasons, such as genotoxic reagents¹. If not repaired promptly, DSBs can result in cell cycle arrest, apoptosis, or senescence. Two major repair pathways—homologous recombination (HR) and non-homologous end joining (NHEJ)—are applied in cells to cope with DSBs^2–4. HR requires homologous sequences as the template for repair; therefore, it mainly has roles in the S and G2 phases of the cell cycle. NHEJ can directly rejoin the broken DNA ends with no to little homology and can function throughout the entire cell cycle^5,6. Because HR copies sequences from homologous templates, it can largely restore the original DNA sequences⁴. NHEJ, however, is more error-prone than HR, and often leaves an ‘information scar’ after repair^2,3. However, many mammalian somatic cells exist in G0 for a large fraction of the time, and NHEJ, therefore, is the predominant pathway for repairing DSBs within cells⁷.

One of the most intriguing and important stages in this major DNA repair pathway is bringing the two DNA ends at a DSB together into physical proximity—a process called synapsis^3,8. The essential components required for NHEJ synapsis have been identified by varied single-molecule techniques^9–15 using blunt-ended DNA substrates. A two-structure synaptic model has been proposed for NHEJ synapsis using both purified NHEJ proteins⁹ and Xenopus crude extracts¹¹. The studies using purified NHEJ proteins show that Ku70-Ku80 (Ku) and XRCC4-DNA ligase IV (X4L4) are sufficient to mediate NHEJ synapsis of blunt ends to form a flexible synaptic (FS) complex, in which the DNA molecules are aligned laterally near the end, and the two DNA ends can loosely search for base pairing within one another; XRCC4-like factor (XLF) and/or Paralog of XRCC4 and XLF (PAXX) facilitates the transition from FS to a close synaptic (CS) state, in which the DNA ends are positioned end-to-end, allowing for ready ligation⁹. Importantly, Ku, X4L4, and XLF/PAXX can mediate CS formation in either a two-step or a one-step manner⁹. Other studies using Xenopus crude extracts suggest a two-stage synapsis mechanism, in which Ku and DNA-PKcs first form a long-range synaptic complex; XLF, non-catalytic X4L4 activity, and DNA-PKcs kinase activity are then required to mediate the transition from long-range synapsis to short-range synapsis. The two DSB ends in the long-range synaptic complex are 115 Å apart, while the ends are in close contact in the short-range synaptic complex and ready for ligation¹¹. Although discrepancies exist for these different studies using different systems^9,11, the core NHEJ proteins—Ku, X4L4, and XLF—are required for NHEJ synapsis and transitions of synaptic complexes, and these proteins mainly function close to the ends of the broken DNAs^8,9,11.

To date, almost all the studies investigating NHEJ synapsis primarily used blunt-ended DNA substrates^9–11, which could allow identification of the essential NHEJ components contributing to synapsis. However, blunt ends are a small portion of the spectrum of DNA end configurations at the DSBs generated endogenously and exogenously^3,16–18. Most DSB ends have diverse configurations and structures, and some even have damaged bases or nucleotides^19,20. Previous studies unveil that the end configurations determine which NHEJ proteins are required for the timely repair of DSBs and the consequences of the repair products²¹. Moreover, damaged bases and mispaired ends were reported to increase the mobility of DNA ends within the synaptic complexes mediated by Ku, X4L4, and XLF proteins, highlighting the importance of end mobility in NHEJ repair¹⁹. While DNA end configurations play important roles in NHEJ repair and may also in the synapsis step, it is unclear if DNA end configurations affect the two structurally different synaptic complexes observed with blunt-ended DNA substrates⁹, and also not clear if the end configurations could determine the choice of NHEJ proteins for DNA synapsis and the structures of formed synaptic complexes, namely the alignments of the two DNA molecules within the synaptic complex. Moreover, it is unclear if different end configurations influence FS which first brings two distant DSBs together, and CS which mainly adjusts the local realignments of two dsDNA at the ends.

Here, using single-molecule FRET (smFRET), we directly evaluate the effects of DNA end configurations on NHEJ synapsis by core NHEJ proteins – Ku, X4L4, and XLF, and investigate the structures of synaptic complexes and the dynamics of synapsis determined by the end configurations. We find that DNA end structures critically influence both the efficiency of NHEJ synapsis and the architecture of the synaptic complexes formed. The presence of 3-nucleotide (nt) microhomology at DNA termini promotes efficient CS complex formation by Ku and X4L4 even in the absence of XLF. While XLF is not essential for close synapsis of ends with 3-nt microhomology, it enhances the efficiency of CS complex. FRET analyses of the synaptic complexes reveal that the single-stranded (ss) and double-stranded (ds) junctions of the overhangs are in close contact in the synaptic complex, while the overhangs themselves form flaps at the DNA ends. These flapped overhangs may facilitate base pairing, thus stabilizing the CS complex mediated by Ku and X4L4, and may also facilitate end processing by NHEJ nucleases and polymerases. Our findings underscore the critical role of DNA end configurations in determining NHEJ synapsis, highlighting their importance across all stages of NHEJ repair.

Results

DNA end configurations influence NHEJ synapsis

Using blunt-ended DNA, we found that Ku70-Ku80 (Ku) together with XRCC4-DNA ligase IV (X4L4) efficiently mediate NHEJ synapsis, forming a flexible synaptic (FS) complex, which is characterized as having a FRET peak around 0.1; and some of the formed synaptic complexes exhibit FRET fluctuations, indicating the mobility of the two dsDNA molecules within the FS complex (Supplementary Fig. 1a–e). XLF promotes the formation of a close synaptic (CS) complex, which is characterized as exhibiting a FRET peak around 0.8–similar to that of the pre-ligated substrates–indicating end-to-end contact of the DNA molecules (Supplementary Fig. 1c, f, g). Dwell-time analysis of the formed synaptic complexes showed lifetimes of 9.9–16.5 s for the FS complex and 29.9 s for the CS complex formed in the presence of XLF (Supplementary Fig. 1h–j). These results are consistent with our previous study⁹, indicating the accuracy of the smFRET system and assay developed here for probing NHEJ synapsis.

The structural features of FS and CS complexes observed for blunt ends suggest that the two ends can be covalently ligated only when a cofactor of X4L4, e.g., XLF, is present in the reaction, which is confirmed by the ensemble ligation assay (Fig. 1a, lane 3). In contrast, compatible overhangs can be directly ligated by Ku plus X4L4 without XLF, with efficiency comparable to that with XLF present (57.5% vs 59.7%) (Fig. 1a, lanes 5 and 6). This indicates that the overhang complementarity can substitute for XLF in producing a ligation-competent CS complex. Moreover, X4L4 alone can ligate compatible overhangs (Supplementary Fig. 1k), further supporting a role for overhang pairing in CS formation.

Fig. 1. DNA end configurations affect NHEJ synapsis.

a Gel analysis of blunt and 3’ overhang ligation by Ku and X4L4 with or without XLF. Similar results were observed across three independent reactions for each substrate. b Schematic of the smFRET assay for 3’ overhang synapsis. Both immobilized and solution duplexes had a terminal 5’ OH and a fluorophore located 4 bp from the dsDNA-overhang junction. c Representative images of the Cy3 channel, the Cy5 channel, and the overlay of two channels for the synapsis of 3’ CCCC and GGGG overhangs mediated by Ku and X4L4. d Normalized synapsis efficiency for blunt ends and 3’ CCCC and GGGG overhangs mediated by Ku and X4L4. Data are presented as mean ± SD from three independent replicates. T test (unpaired, two-tailed) was applied for P value calculation. *P < 0.05 (P = 0.018). Source data are provided as Source Data file.

To directly test the effect of overhangs on NHEJ synapsis, we prepared duplexes with 3’ four-nucleotide complementary overhangs (GGGG/CCCC) for smFRET analysis (Fig. 1b). Both substrates have 5’ OH groups at the ends; therefore, we can focus only on synapsis but not ligation. Ku plus X4L4 can mediate the synapsis of the 3’ GGGG/CCCC overhangs (Fig. 1c and Supplementary Fig. 1I). The lifetimes of synaptic complexes formed by Ku plus X4L4 with these overhangs were similar to those for blunt ends (13.2 vs 16.0 s) (Supplementary Fig. 1h, m), but their synapsis efficiency was ~2.5-fold higher (Fig. 1d). In contrast to blunt ends, where XLF markedly increases synapsis efficiency⁹, XLF had little effect on the synapsis of 3’ GGGG/CCCC overhangs (Supplementary Fig. 1n). Thus, compatible overhangs can compensate for XLF’s role in promoting synapsis, underscoring the influence of DNA end configuration on NHEJ.

Compatible overhangs promote XLF-independent CS formation by Ku plus X4L4

We next examined the structures of synaptic complexes formed by Ku plus X4L4 with 3’ GGGG/CCCC overhangs, focusing on the spatial arrangement of the overhangs within synaptic complexes. FRET trajectories revealed four distinct states with approximate levels of 0.75, 0.55, 0.4, and <0.2 (Fig. 2a and Supplementary Fig. 2a). Corresponding FRET distributions exhibited four obvious peaks, which have the Gaussian-fitted peaks at 0.76, 0.58, 0.39, and 0.13, respectively (Fig. 2b). The highest and lowest peaks match those observed for FS and CS complexes formed with blunt ends in the presence of Ku, X4L4 and XLF (Supplementary Fig. 1f). This indicates that the duplexes within these two types of synaptic complexes formed with 3’ GGGG/CCCC overhangs and with Ku plus X4L4 are organized in a similar way to that in the corresponding complexes with blunt ends and with Ku plus X4L4 plus XLF, namely Ku plus X4L4, without XLF, can promote efficient FS and CS formation when DNA ends have compatible overhangs. These findings suggest that compatible overhangs not only promote CS formation but also facilitate the FS-to-CS transition mediated by Ku and X4L4.

Fig. 2. Compatible overhangs promote CS complex formation by Ku and X4L4.

a Representative single-molecule time traces of donor (green) intensity, acceptor (magenta) intensity, and corresponding E_FRET values (blue) for 3’ GGGG/CCCC overhangs. The light orange dashed lines mark the FRET levels of 0.20, 0.40, 0.55, and 0.75, respectively. b Histograms of E_FRET values of all the synaptic complexes formed with the duplexes with 3’ GGGG/CCCC overhangs. The E values shown were obtained from the Gaussian fits of the corresponding peaks of the synaptic complexes. N = 345 events. c Histograms of E_FRET values of the pre-ligated control substrates, including the blunt ends (8 bp), 3’ GGGG/CCCC overhangs (12 bp), and 3’ GGGGGG/CCCCCC overhangs (14 bp). E values and FWHM from Gaussian fits are 0.76, 0.57, 0.52, respectively, and 0.14, 0.16, 0.11, respectively. Blunt-end data are the same as in Supplementary Fig. 1g. d Possible organizations of the 3’ GGGG/CCCC overhangs within the synaptic complexes mediated by Ku and X4L4. E₁ and E₂ indicate the flexible synaptic (FS) complexes, E₃ indicates the nick-containing or gap-containing close synaptic (CS) complexes, E₄ indicates the junction-contacting close synaptic (CS) complexes. e Schematic of the smFRET assay used to examine synapsis of GGGG/CCCC overhangs containing 8-oxoguanine (8-oxoG) at defined positions. The 8-oxoG within the 3’ GGGG overhang is highlighted in red. f Normalized synapsis efficiency of GGGG/CCCC substrates with or without 8-oxoG in reactions mediated by Ku plus X4L4. Data are presented as mean ± SD from four independent replicates. T test (unpaired, two-tailed) was applied for P value calculation. ns: P > 0.05 (p_{[GGG(8-oxoG)/CCCC vs GGGG/CCCC]} = 0.35, p_{[(8-oxoG)GGG/CCCC] vs [GGGG/CCCC]} = 0.98), ***P < 0.001 (P < 0.0001). g Histograms of E_FRET values for all synaptic complexes formed with GGGG/CCCC substrates containing 8-oxoG. The E values shown correspond to the peak centers obtained from Gaussian fits. Event numbers: N_{[GGG(8-oxoG)/CCCC]} = 408 events, N_{[(8-oxoG)GGG/CCCC]} = 213 events, N_{[(8-oxoG)G(8-oxoG)G/CCCC]} = 181 events. h Fractions of FS and CS complexes formed on GGGG/CCCC overhangs with or without 8-oxoG in the presence of Ku and X4L4. Data are presented as mean ± SD from four independent replicates. T test (unpaired, two-tailed) was applied for P value calculation. ns: P > 0.05 (P = 0.82), *P < 0.05 (P = 0.025), ***P < 0.001 (P = 0.0006). i Dwell times of CS and FS complexes for GGGG/CCCC overhangs with or without 8-oxoG mediated by Ku and X4L4. Data are presented as mean ± SD from four independent replicates. T test (unpaired, two-tailed) was applied for P value calculation. ns: P > 0.05 (P_{[GGG(8-oxoG)/CCCC vs GGGG/CCCC]} = 0.15, P_{[(8-oxoG)GGG/CCCC] vs [GGGG/CCCC]} = 0.30, P_{[(8-oxoG)G(8-oxoG)G/CCCC] vs [GGGG/CCCC]} = 0.92). Note: In the figure, the 8-oxoG incorporated within the 3’ GGGG overhang is highlighted in red. Source data are provided as a Source Data file.

Similar results were obtained for ACTG/CAGT overhangs (Supplementary Fig. 2b–e), including the FRET states of the trajectories (Supplementary Fig. 2d), the FRET distributions of formed synaptic complexes (Supplementary Fig. 2e), and minimal effect of XLF on synapsis efficiency (Supplementary Fig. 2c). These findings indicate that the ability of Ku plus X4L4 to promote CS formation is governed primarily by overhang compatibility rather than by base composition. The A/T-rich TTTT/AAAA overhangs likewise produced similar FRET distributions (Supplementary Fig. 2f), demonstrating that Ku plus X4L4 can support both FS and CS formation even for A/T-rich ends. However, the highest-FRET peak was less frequent than in G/C-rich or ACTG/CAGT overhangs, and in contrast to the G/C-rich ends, XLF enhanced synapsis efficiency for A/T-rich overhangs (Supplementary Fig. 2g), likely reflecting the lower pairing stability of A/T-rich sequences. These observations suggest that base composition modulates synapsis efficiency. Nevertheless, complementarity—rather than base composition—remains the primary determinant of synapsis, with base composition acting only as a secondary modulator.

Two structurally different CS complexes formed for DNA with compatible overhangs

Next, we examined how compatible overhangs are aligned within synaptic complexes to facilitate the formation of, or transition to, the CS complex and to stabilize it. To establish reference FRET values, we prepared several duplex DNA substrates labeled with Cy3 and Cy5 dyes at different distances (Fig. 2c and Supplementary Fig. 2h). These duplexes were generated by pre-ligating either 3’ overhang substrates with different lengths of G/C pairs or blunt-end substrates. Because the Cy3 and Cy5 dyes are positioned 4 bp from the ssDNA–dsDNA junctions on all 3’ overhang substrates, a direct in-line junction–junction contact between the 3’ GGGG/CCCC overhangs results in an 8-bp Cy3–Cy5 separation (Fig. 2d, E₄). Full pairing of the GGGG and CCCC overhangs produces a 12-bp separation with two nicks at the junctions (Fig. 2d, E₃), whereas partial in-line pairing generates a 13–15-bp separation with two gaps between the overhangs (Fig. 2d, E₃). For pre-ligated blunt ends, the Cy3-Cy5 dyes are separated by 8 bp, producing a FRET at 0.76. Therefore, this configuration can serve as a reference for a junction-contacting synaptic complex in which the two duplexes are aligned in-line, and the dsDNA junctions are in direct contact (Fig. 2d, E₄). For overhang substrates to adopt a junction-contacting configuration, the ssDNA overhangs must project outward from the duplex axes to allow dsDNA junctions to touch (Fig. 2d). Pre-ligated compatible GGGG/CCCC and GGGGGG/CCCCCC overhangs have dye separations of 12 bp and 14 bp, respectively. These mimic synaptic complexes in an in-line alignment where the overhangs are base-paired, leaving either two nicks (nick-containing configuration) or two gaps (gap-containing configuration) between them (Fig. 2d, E₃). Because the ends consist of repeated G/C bases, the overhangs can slide relative to each other, allowing partial pairing and transitions between nick-containing and gap-containing states. The corresponding FRET peaks (~0.57 and ~0.52) are close enough to be indistinguishable in our system (Fig. 2c); therefore, we collectively refer to them as the end-pairing configuration.

When 3’ GGGG/CCCC overhangs form synaptic complexes, the most abundant species is the E₄ peak (~0.76; Fig. 2b), matching the pre-ligated blunt-end reference (Fig. 2c). This indicates a junction-contacting configuration in which the ssDNA overhangs flip outward (Fig. 2d, E₄). The next FRET peak, E₃ (~0.58), resembles the pre-ligated GGGG/CCCC and GGGGGG/CCCCCC references, suggesting an end-pairing configuration (including both nick-containing and gap-containing states). Lower FRET peaks, E₂ (~0.39) and E₁ (~0.13), are smaller than the reference duplex with a 14 bp Cy3-Cy5 separation (Fig. 2c), suggesting that the overhangs may be in a gap-containing state with only distal G/C base pairing. However, pairing of only a single G/C base is insufficient for Ku and X4L4 to stabilize an end-pairing CS complex (see next section). Thus, the E₂ and E₁ states represent a looser alignment resembling the FS complex formed by blunt ends (Fig. 2d and Supplementary Fig. 1c).

Since the CS complex formed by blunt ends and by Ku, X4L4, and XLF has close-contact ends, and synaptic complexes of the overhang substrates corresponding to both E₃ and E₄ states involve end-to-end contact, either at the ssDNA overhangs (E₃) or at the ssDNA–dsDNA junctions (E₄), we classify the E₃ and E₄ complexes also as CS complexes. The end-pairing CS complex (E₃) can form de novo (Fig. 2a, top panel), or via transition from either the FS complex (Fig. 2a, bottom panel; Supplementary Fig. 2a, left and middle panels) or the junction-contacting CS complex (E₄) (Supplementary Fig. 2a, left and middle panels). In contrast, the junction-contacting CS complex typically arises from the end-pairing CS complex rather than directly from the FS complex (Fig. 2a and Supplementary Fig. 2a). These results highlight the central role of end-pairing in CS formation mediated by Ku and X4L4.

Compatible ACTG/CAGT overhangs also yield both junction-contacting (E₄) and end-pairing (E₃) CS complexes (Supplementary Fig. 2d, e, i), which excludes the effects of repetitive bases on the formation of CS. Therefore, the results above indicate that at least two structurally distinct CS complexes (junction-contacting and end-pairing) can form for DNA substrates with compatible overhangs. The pairing of the overhangs can compensate for XLF’s role in promoting CS complex formation as required in blunt-end synapsis. The (potential) pairing of the end overhangs either as in the junction-contacting CS complex or the end-pairing (nick-containing and gap-containing) CS complex can stabilize the CS complex formed by Ku plus X4L4.

We further tested whether the common oxidative lesion 8-oxoguanine affects synapsis mediated by Ku plus X4L4. 3’ GGGG/CCCC substrates containing one or two 8-oxoguanines at defined positions were designed (Fig. 2e). A single lesion did not impair synapsis efficiency, whereas two lesions caused only a modest reduction (Fig. 2f and Supplementary Fig. 3). Neither one nor two lesions disrupted CS formation or stability (Fig. 2g, i). Thus, 8-oxoguanine has minimal impact on synapsis—including synapsis efficiency and CS formation—mediated by Ku plus X4L4, consistent with previous findings that NHEJ can tolerate 8-oxoguanine within the overhangs at DNA ends^19,22.

A minimum of three complementary base pairs are required for efficient CS formation by Ku and X4L4

We next asked how many complementary base pairs are needed to promote CS by Ku and X4L4. To address this, we tested DNA substrates with compatible ends ranging from one (G/C) to three (GGG/CCC) base pairs (Fig. 3a). Single-molecule images and trajectories confirmed that Ku and X4L4 can mediate synapsis of all these substrates (Fig. 3b–e). However, synapsis of ends with one (G/C) or two (GG/CC) base pairs yielded primarily low FRET signals (<0.5) (Fig. 3c, d), similar to blunt ends (Supplementary Fig. 1e). Corresponding FRET distributions of all the synaptic complexes for G/C as well as GG/CC overhangs are also similar to that of blunt ends. The main peaks of FRET distributions are around 0.11 and 0.09, respectively (Fig. 3f, g), which are substantially lower than those of pre-ligated overhangs (Supplementary Fig. 4a). Thus, Ku and X4L4 are not able to proceed beyond FS complexes for substrates with ≤2 base pairs, suggesting that two base pairs at ends do not provide sufficient energy for the FS-to-CS transition.

Fig. 3. Three complementary base pairs are required for efficient CS formation by Ku and X4L4.

a Schematic of the smFRET assay for 3’ overhangs (1–3 nt). Both the immobilized and the solution duplexes had a terminal 5’ OH and a fluorescence dye located 4 bp away from the ds-ss junction. b Representative images of Cy3 channel, Cy5 channel, and overlay of two channels for synapsis of duplexes with 3’ GGG/CCC overhangs. c–e Representative single-molecule time traces of donor (green) intensity, acceptor (magenta) intensity and corresponding E_FRET values (blue) for synapsis of duplexes with 3’ overhangs. The light orange dashed lines mark the FRET levels of 0.20, 0.40, 0.55, and 0.75, respectively. f–h Histograms of E_FRET values of all the synaptic complexes formed with 3’ overhang substrates mediated by Ku and X4L4. The E value shown was obtained from a Gaussian fit of the highest peak of the synaptic complexes. f N = 215 events; g N = 191 events; h N = 392 events. i Fractions of FS and CS complexes mediated by Ku and X4L4. Data are represented as mean ± SD from the following numbers of biological replicates: n_{[blunt/blunt]} = 8, n_[G/C] = 5, n_[GG/CC] = 6, n_[GGG/CCC] = 8, and n_[GGGG/CCCC] = 5. j Normalized synapsis efficiency across 3’ overhang substrates. Data are represented as mean ± SD from the following numbers of biological replicates: n_{[blunt/blunt]} = 5, n_[G/C] = 5, n_[GG/CC] = 6, n_[GGG/CCC] = 8, n_[GGGG/CCCC] = 5. T test (unpaired, two-tailed) was applied for P value calculation. ns: P > 0.05 (P = 0.90), ***P < 0.001 (P < 0.0001). k Schematic of the smFRET assay used to examine synapsis of GGGTT/CCCTT overhangs mediated by Ku plus X4L4. l Histograms of E_FRET values for all the synaptic complexes formed with 3’ GGGTT/CCCTT overhangs mediated by Ku plus X4L4. The E value shown was obtained from a Gaussian fit of the highest peak of the synaptic complexes. N = 182 events. m Fractions of FS and CS complexes formed on GGGTT/CCCTT overhangs mediated by Ku plus X4L4. Data are presented as mean ± SD from five independent replicates. T test (unpaired, two-tailed) was applied for P value calculation. ***P < 0.001 (P < 0.0001). n Histogram and corresponding exponential fits of synapsis dwell times for CS events mediated by Ku plus X4L4. N = 40 events. Only the events with both start and end time points within the detection time window were included. Source data are provided as a Source Data file.

By contrast, overhangs with three base pairs (GGG/CCC) produced high FRET traces (Fig. 3e), with FRET distributions similar to those of GGGG/CCCC substrates (Figs. 2b, 3h). These results indicate that Ku and X4L4 can promote efficient CS complexes formation, including junction-contacting complex (E₄) and end-pairing complex (E₃), for GGG/CCC overhangs (Supplementary Fig. 4b), thus at least three base pairs between overhangs are needed to provide the energy for efficient FS-to-CS transition and to stabilize the formed CS. The quantification of FS and CS also showed that CS occurs only for overhangs with ≥3 base pairs (Fig. 3i), thus indicating a minimum of three base pairs between ends is required for efficient CS formation by Ku and X4L4. Although synapsis efficiency did not change significantly between one- and two-base pair overhangs, efficiency increased markedly when complementarity reached four base pairs (Fig. 3j). This indicates that extending microhomology not only promotes synapsis efficiency but also favors CS formation, likely through overhang pairing that stabilizes the transition from FS to CS. Nevertheless, the stability of the CS complex as reflected by the duration exhibited insignificant differences between GGG/CCC and GGGG/CCCC substrates (Supplementary Fig. 4c), suggesting that three base pairs are sufficient to form a stable CS.

To further assess the contribution of paired overhangs, we varied the length of the overhangs on immobilized substrates (0–3 Gs) while maintaining four complementary Cs on the solution substrate (Supplementary Fig. 4d). The results showed that an obvious CS complex corresponding to the FRET peak of E₄ were observed only when three complementary Gs were present on the immobilized end (Supplementary Fig. 4e–i), and the dwell times of CS complexes were comparable between GGG/CCC and GGG/CCCC overhangs (Supplementary Fig. 4c, j). We also examined whether Ku and X4L4 can promote CS formation when the three-base-pair microhomology is embedded within longer, partially incompatible overhangs (GGGTT/CCCTTT; Fig. 3k–n and Supplementary Fig. 5). This scenario is relevant for understanding synapsis prior to end processing, such as Artemis-mediated trimming. Ku and X4L4 did promote CS formation on GGGTT/CCCTTT overhangs (Fig. 3l, m); however, both CS stability and formation efficiency were reduced relative to GGG/CCC (dwell: 4.2 s vs. 10.3 s; CS fraction: 19.6% vs. 34.8%; Fig. 3i, m, n and Supplementary Fig. 4c). These results show that three embedded complementary base pairs are sufficient for CS formation, but incompatible flanking sequences diminish the efficiency and stability of the complex.

These findings demonstrate that a minimum of three complementary base pairs is required for Ku and X4L4 to promote efficient and stable CS formation.

XLF promotes CS formation from overhangs with limited complementarity

XLF, an accessory factor of X4L4²³, promotes ligation by facilitating the FS-to-CS transition⁹. To test its role in overhang synapsis, we applied the same DNA substrates as in Fig. 3 (Fig. 4a). The results showed that increasing the number of complementary base pairs enhanced synapsis in the presence of XLF (Supplementary Fig. 6a, b), indicating that base pairing contributes to synapsis even when XLF is present. For single-base overhangs (G/C), XLF enabled formation of high FRET complexes with a peak at 0.68 (Fig. 4b), comparable to pre-ligated G/C overhangs (Supplementary Fig. 4a) and within the range of blunt-end CS (Fig. 2c). Similarly, GG/CC overhangs displayed high FRET peaks in the presence of XLF (Fig. 4c). Notably, the highest FRET peak for GG/CC (0.68) was larger than that of pre-ligated GG/CC (0.64) (Fig. 4c and Supplementary Fig. 4a), consistent with formation of junction-contacting CS complexes. These results demonstrate that XLF is required for CS formation when microhomology is limited to fewer than three base pairs. For longer overhangs (GGG/CCC and GGGG/CCCC), the FRET distributions exhibited predominant high FRET peaks of 0.72–0.73 (Fig. 4d, e), which are close to that of the blunt-end CS complex (Fig. 2c). This suggests that XLF further promotes junction-contacting CS complexes, likely through interactions with Ku and X4L4 as previously reported⁹. Moreover, across all substrates, XLF increased the proportion of CS complexes relative to reactions without XLF (Fig. 4b–f vs Fig. 3f–i), with CS abundance correlating with overhang length (Fig. 4f). The results above demonstrate that base pairing between the overhangs and XLF contributes to CS complex formation.

Fig. 4. XLF promotes CS formation.

a Schematic of the smFRET assay for synapsis of 3’ overhangs mediated by Ku, X4L4, and XLF. b–e Histograms of E_FRET values of all the synaptic complexes formed with 3’ overhangs in the presence of Ku, X4L4, and XLF. The E value shown was obtained from a Gaussian fit of the highest peak of the synaptic complexes. b N = 160 events; c N = 243 events; d N = 248 events; e N = 241 events. f Fractions of FS and CS complexes mediated by Ku, X4L4, and XLF for 3’ overhangs. Data are presented as the mean ± SD from independent biological replicates, with sample sizes as follows: n_{[blunt/blunt]} = 6, n_[G/C] = 3, n_[GG/CC] = 3, n_[GGG/CCC] = 3, and n_[GGGG/CCCC] = 3. T test (unpaired, two-tailed) was applied for P value calculation. ns: P > 0.05 (P_{[GGGG/CCCC vs GG/CC]} = 0.11, P_{[GGGG/CCCC] vs [GGG/CCC]} = 0.10), *P < 0.05 (P = 0.024), ***P < 0.001 (P < 0.0001). g Dwell times of CS complexes stimulated by Ku, X4L4, and XLF. Data are presented as the mean ± SD from independent biological replicates, with sample sizes as follows: n_{(Ku+X4L4: GGG/CCC)} = 4, n_{(Ku+X4L4: GGGG/CCCC)} = 5, n_{(Ku+X4L4+XLF: GG/CC)} = 4, and n = 3 for all other samples. T test (unpaired, two-tailed) was applied for P value calculation. ns: P > 0.05 (P_{[GGG/CCC: Ku+X4L4 vs Ku+X4L4+XLF]} = 0.079, P_{[GGGG/CCCC: Ku+X4L4 vs Ku+X4L4+XLF]} = 0.067). h Histograms of E_FRET values for all the synaptic complexes formed with 3’ GGGTT/CCCTT overhangs mediated by Ku plus X4L4 in the presence of XLF. The E value shown was obtained from a Gaussian fit of the highest peak of the synaptic complexes. N = 562 events. i Fractions of FS and CS complexes formed on GGGTT/CCCTT overhangs mediated by Ku, X4L4, and XLF. Data are presented as mean ± SD from five independent replicates. T test (unpaired, two-tailed) was applied for P value calculation. ***P < 0.001 (P = 0.0002). j Histogram and corresponding exponential fits of synapsis dwell times for CS events mediated by Ku plus X4L4 in the presence of XLF. N = 68 events. Only the events with both start and end time points within the detection time window were included. Source data are provided as a Source Data file.

Dwell-time analysis revealed no significant difference between CS complexes formed with or without XLF, or among different overhang lengths (Fig. 4g). The results suggest that base pairing between overhangs can supplement the bridging role of XLF in promoting CS formation, and both aspects contribute to CS complex formation, thus leading to a higher amount of CS for overhangs with longer microhomologies (Fig. 4f). In addition, XLF substantially enhanced both the formation and stabilization of CS complexes for GGGTT/CCCTT overhangs (Supplementary Fig. 5b and Fig. 4h–j), indicating that XLF also plays a key role in synapsis of partially compatible overhangs.

Together, these results above demonstrate that XLF can promote CS complex formation across substrates, especially for overhangs with limited microhomology.

DNA end chemistry influences CS formation

It has been reported that synapsis can also be modulated by end chemistry, particularly the presence of 5’ phosphate groups (5’ P), which facilitate synapsis and stabilize complexes formed by Ku, X4L4, and XLF²⁴. To examine this in the context of overhangs by Ku and X4L4, we prepared DNA substrates with 5’ P at both ends (Fig. 5a). The results showed that synapsis efficiency increased significantly for overhangs longer than one base pair when 5’ P was present (Fig. 3j and Supplementary Fig. 7a, b).

Fig. 5. DNA end chemistry influences CS formation mediated by Ku and X4L4.

a Schematic of the smFRET assay for synapsis of phosphorylated blunt and 3’ overhang ends mediated by Ku and X4L4. All duplexes were synthesized with a 5’ PO₄. b–d Histograms of E_FRET values of all the synaptic complexes formed with phosphorylated duplexes, Ku, and X4L4. The E value shown was obtained from a Gaussian fit of the highest peak of the synaptic complexes. b N = 222 events; c N = 217 events; d N = 283 events. e Fractions of FS and CS complexes mediated by Ku and X4L4 for phosphorylated duplexes. Data are presented as the mean ± SD from independent biological replicates, with sample sizes as follows: n_{(blunt/blunt)} = 4, n_(G/C) = 4, n_(GG/CC) = 5, and n_(GGG/CCC) = 4. T test (unpaired, two-tailed) was applied for P value calculation. ***P < 0.001 (P < 0.0001). f Ratio of ligated synaptic complexes to total complexes after 10 min reaction. In all, 2 M NaCl was used to wash the reaction chamber to release any noncovalent association of duplexes within the synaptic complexes. Data are presented as mean ± SD from three independent replicates. T test (unpaired, two-tailed) was applied for P value calculation. ns: P > 0.05 (P = 0.29), ***P < 0.001 (P_{[blunt/blunt vs GG/CC]} = 0.0002, P_{[blunt/blunt vs GGG/CCC]} = 0.0005). Source data are provided as a Source Data file.

With 5’ phosphorylated G/C overhangs, FRET distributions showed only low peaks (E₁ = 0.11 and E₂ = 0.3), suggesting the formation of FS complexes (Fig. 5b). In contrast, 5’ phosphorylated GG/CC overhangs yielded three dominant peaks (E₂ = 0.16, E₃ = 0.51, and E₄ = 0.72) (Fig. 5c), suggesting the formation of FS (E₂), nicking-containing CS (E₃), and junction-contacting CS (E₄) complexes. A single 5’ phosphate at one end also promoted CS formation to a similar extent as two phosphorylated ends (Supplementary Fig. 7c–h), although dual phosphorylation provided a modest additional enhancement for GG/CC overhangs (Fig. 5e and Supplementary Fig. 7h). The formation of CS complexes for 5’ phosphorylated GG/CC overhangs indicates that both base pairing (two nucleotides) and 5’ P act cooperatively to promote CS formation. Because GG/CC overhangs without 5’ P primarily formed FS (Fig. 3g) and G/C overhangs even with 5’ P failed to form CS (Fig. 5b). The interactions between 5’ P and adenylated Ligase IV^25,26 may help stabilize and transition intermediates toward CS. 5’ P also promoted CS formation with GGG/CCC substrates (Figs. 3h–i, 5d, e). Moreover, duplexes synapsed within CS complexes from GG/CC and GGG/CCC overhangs were resistant to 2 M NaCl (Fig. 5f), indicating covalent ligation of the duplexes by X4L4. This was also confirmed by ensemble ligation assays, which showed successful ligation of these overhangs by Ku and X4L4 (Supplementary Fig. 8). These results reveal that CS complexes formed with overhangs by Ku and X4L4 are ligation-competent. The ligation likely occurs directly within the nick-containing CS complexes or following transitions from junction-contacting to nick-containing CS states.

Together, these findings show that DNA end chemistry, specifically 5’ phosphorylation, enhances both synapsis efficiency and the formation of ligation-competent CS complexes.

Discussion

Bringing two broken dsDNA ends together is the first and most critical step for the repair of most double-strand DNA breaks and is essential for their eventual ligation. Core NHEJ proteins—Ku, X4L4, and XLF—have previously been shown to mediate synapsis of blunt ends, forming two distinct types of synaptic complexes—flexible synaptic (FS) and close synaptic (CS) complexes^9,20. Ku and X4L4 are essential for the initial synapsis of blunt ends, but they only support the formation of FS complexes, which are ligation-incompetent. XLF must be present for mediating the transition of the FS complex to the ligation-competent CS complex. However, our previous work did not establish whether, or how, the end sequences influence synapsis and thereby affect the ligation outcome. Although blunt ends are useful for identifying the essential components required for NHEJ synapsis^9–11, they do not reflect the diversity of broken DNA ends encountered in vivo – especially those generated by physical or oxidative damage^3,19,20. Ensemble ligation studies have shown that DNA end configuration can dictate the specific requirements of NHEJ components for end joining^2,21. Here, we demonstrate that DNA end configuration also has critical roles in determining NHEJ synapsis (Fig. 6), contributing to the mechanistic diversity of NHEJ synapsis. The existence of multiple synapsis pathways likely equips the NHEJ machinery to efficiently handle a wide range of end structures, thereby ensuring timely DSB repair and genomic stability.

Fig. 6. End-configuration-dependent model of NHEJ synapsis.

Following DSB formation, DNA ends with varied sequences can arise. Ku binds broken ends and recruits X4L4 to mediate flexible synaptic (FS) complexes, in which ends are loosely aligned, permitting a search for microhomology. Pairing of ≥3-nt microhomologies stabilizes the FS-to-CS transition, yielding nick- or gap-containing CS complexes that can further transition to stable junction-contacting CS complexes. In these junction-contacting CS complexes, ss–ds junctions are closely apposed while overhangs extend outward, allowing additional processing or microhomology pairing. These CS complexes are ligation-competent intermediates that transition to nick-containing CS complexes for end joining. When microhomology is <3 nt, FS-to-CS transition requires XLF. Note that when microhomology is 2 nt but with 5’ phosphorylated ends, the interaction of DNA ligase IV with the phosphate group can also mediate the FS-to-CS transition without the presence of XLF. This model highlights the role of end configuration in dictating NHEJ synapsis. MH microhomology, 5’ P 5’ phosphorylation.

Following DSB formation, varied end sequences can appear. Ku rapidly binds to the broken ends and recruits X4L4. Together, these proteins mediate the synapsis of two ends to form FS complexes, in which the two DNA molecules are laterally aligned near their ends, allowing the ends to loosely probe one another for base pairing to identify microhomologies⁹ (Fig. 6). When microhomologies longer than two nucleotides are present, their pairing provides sufficient energetic favorability to drive the transition from FS to CS, even when XLF is absent (Fig. 6). However, when microhomology is minimal (e.g., <3 nt), this transition relies heavily on XLF (Fig. 6).

Interestingly, in most CS complexes formed with overhangs, the DNA ends are typically arranged in an in-line configuration, and the ssDNA–dsDNA junctions are in close contact, with the overhangs flipping outward—forming a junction-contacting CS complex (Fig. 6). This configuration may allow further sampling and stabilization of microhomology. In addition, the flapped overhangs provide spatial flexibility for processing incompatible sequences—for example, by Artemis²⁷—without allowing the ends to drift apart. Artemis, through its endonuclease activity, can cleave incompatible flaps near ssDNA–dsDNA junctions²⁷ within the synaptic complexes. The resulting compatible overhangs can then efficiently promote and stabilize CS (Fig. 3) and ultimately enable end ligation by Ku and X4L4²⁷. This could also provide a mechanism to explain how incompatible overhangs are processed within the close synaptic complex, as has been proposed elsewhere²⁸, although the precise coupling mechanism of end processing and synapsis is still unclear and under investigation^29,30. Since the formed CS complexes are ligation-competent, the junction-contacting CS complex must eventually transit into a nick-containing CS complex to enable covalent end ligation (Fig. 6).

XLF has been reported to stimulate the activity of X4L4 for blunt ends through promoting the formation of the ligation-competent CS complexes^8,9,11,15. Here, we further show that XLF also facilitates CS complex formation for DNA with overhangs, particularly when microhomology between overhangs is less than 3 nt—conditions under which Ku and X4L4 alone cannot efficiently drive the FS-to-CS transition (Figs. 3i, 4f). This highlights a specific role for XLF in promoting ligation when the ends lack sufficient pairing energy. Ensemble ligation assays confirmed the single-molecule results (Supplementary Fig. 8), demonstrating that XLF becomes critical only when end sequences cannot compensate for its function. These findings are consistent with previous studies showing that XLF is essential for some DNA substrates but dispensable for others^18,21,29,31. Together, these results underscore the pivotal role of end sequence in determining the pathway and requirements for NHEJ synapsis.

In addition to the end sequence, end chemistry has been reported to modulate NHEJ synapsis. The presence of 5’ phosphate (5’ P) has been reported to enhance both synapsis efficiency and stability of DNA end pairing²⁴. Our results extend this finding by showing that 5’ P also facilitates the FS-to-CS transition in the absence of XLF. This may be due to an interaction between the 5’ P and the adenylated intermediate of Ligase IV, as previously suggested^25,26. However, 5’ P alone is not sufficient to replace the stabilizing effect of base pairing; for example, two complementary nucleotides are still required for efficient CS formation (Figs. 5, 6). These results suggest that 5’ P plays a modulatory, but not dominant, role in promoting synapsis²⁴.

The observation that the end sequence determines the choice of NHEJ synapsis pathway is in agreement with previous in vitro and in vivo studies showing that end configurations dictate NHEJ protein requirements during end joining^18,20,21. The finding that minimal end microhomologies can support CS complex formation by Ku and X4L4 complements earlier studies showing that CS formation can also be driven by other core NHEJ components^8,9, as well as by X-family DNA polymerases such as polymerase mu (pol μ)³² and polymerase lambda (pol λ)³³. These results collectively highlight the flexibility and complexity of NHEJ synapsis. The mechanistic diversity of NHEJ allows it to accommodate a broad range of broken end structures, including those arising during genome editing, V(D)J recombination, and class switch recombination, thereby ensuring high efficiency in repairing DSBs and maintaining genome integrity.

Methods

DNA substrates

The oligonucleotides used in this study were synthesized and purified by Sangon Biotech (Shanghai, China). Their sequences are listed in Supplementary Data 1. Duplexes were prepared by annealing corresponding oligonucleotides (Supplementary Table 1) in the annealing buffer of 20 mM Tris-HCl (pH 8.0) and 100 mM NaCl. The double-stranded DNA (dsDNA) substrates immobilized on the slide include 5’OH/P-BZ15/HC120, 5’OH/P-BZ15/HC124, 5’OH/P-BZ15/HC123, 5’OH/P-BZ15/HC122, 5’OH/P-BZ15/HC116, 5’OH/P-BZ15/CZ70, 5’OH/P-BZ15/BZ80, 5’OH/P-BZ15/CZ72, 5’OH/P-BZ15/CZ93, 5’OH/P-BZ15/CZ94, 5’OH/P-BZ15/CZ95, 5’OH/P-BZ15/CZ87 and 5’OH-BZ15/HC119, in which the oligonucleotides HC120, HC124, HC123, HC122, HC116, HC119, CZ70, CZ72, CZ93, CZ94, CZ95 and CZ87 contain a biotin at their 5’ end for immobilization. The duplexes used in solution include 5’OH/P-BZ35/(BZ52 + BZ24), 5’OH/P-BZ35/(BZ71 + BZ24), 5’OH/P-BZ35/(BZ55 + BZ24), 5’OH/P-BZ35/(BZ70 + BZ24), 5’OH/P-BZ35/(BZ36 + BZ24), 5’OH/P-BZ35/(BZ81 + BZ24), 5’OH/P-BZ35/(CZ69 + BZ24), 5’OH/P-BZ35/(CZ71 + BZ24), 5’OH/P-BZ35/(CZ92 + BZ24) and 5’OH/P-BZ35/(CZ89 + BZ24). These solution duplexes contain a loop at one end to let synapsis only happen at the free end. 5’OH/P-BZ35 and BZ24 first annealed with BZ52, BZ71, BZ55, BZ70, BZ36, CZ69, CZ71, CZ92 and CZ89, respectively, to form a loop- and nick-contained duplex. The nicks were then sealed by T4 DNA ligase (Sangon Biotech, Shanghai) at 15 °C overnight. Positive control substrates were prepared by ligating 5’P-BZ15/HC120 with 5’P-BZ35/(BZ52 + BZ24), 5’P-BZ15/CZ72 with 5’P-BZ35/(CZ71 + BZ24) and 5’P-BZ15/HC116 with 5’P-BZ35/(BZ36 + BZ24), respectively. The ligation products were then resolved by 10% urea-denaturing PAGE gels and the target products on the gels were then sliced and further purified.

Protein expression and purification

Ku70/80, X4L4, and XLF were expressed and purified following the previously described protocols⁹ (Supplementary Fig. 1a). Briefly, Ku70/80 was expressed in High Five insect cells (Invitrogen, B855-02) and purified using Ni-NTA affinity beads (Union-Biotech, Shanghai), HiTrap Heparin HP column (Cytiva, Sweden AB, 17040601), HiTrap Capto Q ImpRes anion exchange column (Cytiva, Sweden AB, 17547055), and Superdex 200 Increase column (Cytiva, Sweden AB, 28990944). Ligase IV was co-expressed with XRCC4 in High Five insect cells and purified using Ni-NTA affinity beads, anion exchange column, and Superdex 200 column. XLF was expressed in 293 T cells (ATCC, CRL-3216) by transfecting the cells with XLF cDNA contained pcDNA6 plasmid (pcDNA6/huXLF-Myc-His). After expression, XLF was purified via Ni-NTA affinity chromatography and anion exchange chromatography. The concentrations of purified recombinant proteins were measured by a 12% SDS-PAGE using BSA as the standard. Purified proteins were aliquoted and stored at −80 °C after flash-frozen using liquid nitrogen.

Ensemble NHEJ assay

The ensemble NHEJ reaction mixture included 40 nM each of the two dsDNA substrates, 1 mM ATP, 2 mM DTT, 50 nM Ku70/80, and 100 nM X4L4, with or without 100 nM XLF. The ligation buffer consisted of 25 mM Tris-HCl (pH 8.0), 75 mM KCl, and 10 mM MgCl₂. Samples were incubated at 37 °C for 1 h, followed by phenol extraction and ethanol precipitation. Products were separated by 12% native PAGE and visualized using a ChemiDoc MP Imaging System with parameters for Cy5 excitation and imaging.

Slide and coverslip modification

Slides and coverslips were modified with biotin-PEG (Laysan Bio, biotin-PEG-SVA-5000) and mPEG (Laysan Bio, mPEG-SVA-5000). They were first soaked in 1 M KOH for 20 min, rinsed with purified water, and treated with a 75% H₂SO₄/25% H₂O₂ solution for 20 min. After rinsing with purified water and washing with methanol, slides and coverslips were silylated with a methanol solution containing 2% (v/v) APTES (Millipore Sigma, 281778) and 5% (v/v) acetic acid. After rinsing and drying, they were PEGylated overnight at room temperature with mPEG or a mixture of mPEG and biotin-PEG. Finally, they were washed, dried, and stored at −20 °C in a sealed container until use.

Single-molecule setup

The single-molecule FRET (smFRET) experiments were carried out on the Olympus IX83P2ZF inverted microscope equipped with the 532 nm (COHERENT, OBIS 532 nm LS 50 mW) laser. The output laser was collimated and reflected by a mirror (Chroma ZT405/488/532/640rpc-XT) into an oil immersion objective (Olympus, 100×, NA = 1.5) to achieve the TIRF illumination. The laser power was set at 1.5 mW before the objective lens. A trigger box, which is based on the UNO board from Arduino and is provided by FluoCa Scientific (Shanghai, China), was applied to synchronize the laser excitation and the camera exposure during imaging. The Cy3 and Cy5 emission wavelengths passed through a bandpass filter (Chroma ZET405/488/532/642 m) mounted in the microscope, and were further separated by a dichroic beam splitter (Semrock FF640-FDi01-25×36) mounted in the OptoSplit II (CAIRN RESEARCH). The Cy3 and Cy5 emissions were finally imaged on the two halves of a back-illuminated sCMOS camera (Teledyne Photometrics, Prime 95B) after passing the bandpass filters of ET600/50 m (Chroma) and ET700/75 m (Chroma), respectively. The smFRET system was controlled, and the images were captured both with the Micro-Manger 2.0 software^34,35.

Single-molecule imaging

The smFRET strategy used for the NHEJ synapsis investigation is similar as previously described^9,32. Briefly, two duplexes (74–85 bp) with varied end configurations were prepared with Cy3 or Cy5-labeled at a distance of 4 bp from the double-stranded (ds) terminus and close to the single-stranded (ss) overhang. The Cy5-labeled dsDNA is immobilized on the slide, the Cy3-labeled dsDNA is free in solution and has a loop at the opposite side to the Cy3 dye. The simultaneous appearance of Cy3 and Cy5 signals indicates the occurrence of synapsis, and the changes in FRET (E_FRET) signals reflect the dynamic interactions between two dsDNA ends.

The 5’OH/P-dsDNA was immobilized on the surface of the coverslips by a biotin-streptavidin-biotin strategy at a concentration of 300 pM. 10 nM loop-containing and Cy3-labeled 5’OH/P-dsDNA was used in the solution. Unless otherwise specified, the reaction buffer in this study contained 20 mM Tris-acetate pH 7.5, 75 mM KCl, 10 mM MgCl2, 3% glycerol, 1 mg/ml BSA, 1 mM DTT, 0.5 mM ATP, 0.25% Tween 20, 2 mM Trolox (Millipore Sigma, 238813) and an oxygen-scavenging system, which contains 0.8 % (w/v) D-glucose, 1 mg/ml glucose oxidase (Millipore Sigma, G2133-50KU), and 0.04 μg/ml catalase (Millipore Sigma, C3556). The final concentrations of the proteins were 50 nM Ku, 100 nM X4L4, and 100 nM XLF. Proteins and looped dsDNA were added to the reaction buffer, mixed, and immediately injected into the imaging chamber. Images were captured immediately after sample injection. All reactions were carried out at room temperature (22 °C). The exposure time of the CCD camera was set to 100 ms and each movie lasted approximately 2 min, with at least three movies taken for each sample. Each sample was repeated at least three times in this study. In all, 532 nm laser was used constantly to excite FRET pairs during each ROI imaging session. In all, 532 nm laser cannot directly excite the Cy5 dye, the Cy5 dye only brightens when it is exposed to energy transferred from the excited Cy3 dye. The appearance of both Cy3 and Cy5 signals or the FRET signal indicates the occurrence of synapsis. The emitted fluorescence signals of Cy3 and Cy5 were split into two channels using a splitter (FF640-FDi01-25×36, Semrock). The Cy3 and Cy5 imaging channels were registered using the BZ45/ZhaoBZ49 dsDNA probe, which contains Cy3 and distance-rational Cy5 dyes.

smFRET data analysis

FRET pair extraction and further data analysis were similar to our previous study⁹. Briefly, FRET pairs were searched and filtered by scanning every 20 frames of the movie using the iSMS v2.01 software built-in Matlab^36,37. Since the slide surface was not completely blocked, Cy3-dsDNA in solution may adsorb non-specifically to the slide surface, resulting in some spots of Cy3 signals on the images (called Cy3-only spots). Only molecules with paired Cy3 and Cy5 signals were selected, and synapsis events were considered as molecules with paired Cy3 and Cy5 signals. All identified trajectories were then exported, and the data were further analyzed using programs written in MATLAB, as described previously⁹. The synapsis efficiency was measured from the number of events obtained from the reaction of 50 nM Ku and 100 nM X4L4 or the reaction of 50 nM Ku, 100 nM X4L4 and 100 nM XLF as specified in the figures. Synapsis efficiency was normalized by the number of synaptic molecules from the synapsis of blunt ends under the same condition. All molecules of each sample were combined and used to plot the E_FRET histograms, and all E_FRET histograms were fitted with the Gaussian distribution functions.

FS and CS complexes were distinguished based on FRET levels observed in synapsis events. Classification was guided by the full width at half maximum (FWHM) of FRET distributions obtained from pre-ligated DNA substrates. Specifically, the pre-ligated blunt-end substrate—with an 8-nt Cy3–Cy5 separation—was used as a reference to identify junction-contacting CS complexes. Because 2-nt microhomology is insufficient for stable CS formation by Ku and X4L4, the highest FRET cutoff (0.66) derived from pre-ligated 3’ GGG/CCC overhangs (peak FRET = 0.60; FWHM = 0.11; Supplementary Fig. 4a) was also adopted as a threshold for defining junction-contacting CS. Thus, synapsis events with FRET ≥ 0.66 were classified as junction-contacting CS complexes. This cutoff is consistent with the lowest cutoff (0.69) observed for pre-ligated blunt ends (Supplementary Fig. 1g or Fig. 2c), which also defines junction-contacting CS complexes. The gap-containing CS complex formed by 3’ GGGG/CCCC overhangs was expected to yield the lowest CS FRET due to the 13-nt Cy3–Cy5 separation. This configuration was modeled using pre-ligated 3’ GGGGG/CCCCC overhangs (Supplementary Fig. 2h), which gave a cutoff of 0.48 at FWHM. Accordingly, synapsis events with FRET > 0.48 were categorized as CS (including gap-containing, nick-containing, and junction-contacting CS), whereas events with FRET ≤ 0.48 were classified as FS. This 0.48 threshold is further supported by FRET histograms of 3’ G/C and 3’ GG/CC overhangs (Fig. 3f, g), which form only FS complexes with Ku and X4L4 and predominantly exhibit FRET below 0.48. Following classification, the formation efficiency and dwell times of FS and CS complexes were calculated for each DNA substrate set.

Synapsis time analysis

The dwell-time analysis was based on the time window from the appearance to the disappearance of each FRET pair of molecules. All the dwell-time histograms and the corresponding exponential fits corresponded only to synapsis events that begin and end within the detection time window. Distributions were fitted with single-exponential decay functions.

Quantification and statistical methods

The Gaussian fitting of the E_FRET distribution and the exponential fitting of the dwell time distribution were performed using OriginLab Origin 2024b. The statistical analysis was performed using the unpaired two-tailed t test built-in Excel and statistical significance was considered when the P value was less than 0.05. Statistical analysis was performed with at least three independent experiments.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization, methodology, and validation: D.L. and B.Z.; formal analysis and investigation: C.Z. (Chenyang Zhang), Y.J., D.L., and B.Z.; writing—original draft, review and editing: C.Z. (Chenyang Zhang), W.J., W.L. (Weilin Li), R.F., K.Q. W.L. (Wantao Li), C.Z. (Chao Zhao), D.L., and B.Z.; resources, supervision, and project administration: D.L. and B.Z.; funding acquisition: C.Z. (Chenyang Zhang), W.J., R.F., D.L., and B.Z. All authors contributed to the revision of the manuscript and approved the final version.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68417-0.