DNA double-strand breaks (DSBs) are repaired by Ku-dependent non-homologous end-joining (NHEJ) or DNA end resection-dependent homologous recombination (HR). Previously, it was shown that short-range resection by MRX-pSae2 can remove Ku from DSBs and initiate resection, showing flexibility in DSB repair pathway choice. However, which NHEJ intermediates can be disrupted by resection factors remains unknown. Using reconstituted reactions with purified S. cerevisiae proteins, we show here that short-range resection by MRX-pSae2 is not directly inhibited by any of the core NHEJ factors. Instead, we find that MRX-pSae2 is inhibited structurally. DNA synapsis, a dynamic NHEJ intermediate that bridges DSBs, strongly suppresses resection. Non-ligatable ends are refractory to MRX-pSae2 once synapsis formed, and NHEJ assemblies with catalytic mutants of DNA ligase IV that permit synapsis but do not allow ligation likewise inhibit resection. We further demonstrate that removal of Ku from DSBs requires RPA, which interacts with MRX and pSae2 to synergistically stimulate MRX 3′-5′ exonuclease activity, generating free 3′ overhangs. Our data support a model in which DNA synapsis temporarily prevents resection, allowing NHEJ to execute ligation. If ligation fails and synapsis dissociates, MRX-pSae2 and RPA promote resection that removes non-productive NHEJ assemblies and channels DSB repair toward HR. The data help explain the suppression of NHEJ by RPA observed in previous genetic experiments.

In the uploaded files, we present the source data of the radioactive assays presented in the paper, as well as mass photometer, single-molecule magnetic tweezer, and nanofluidics data from the main and supplementary figures. The uploaded files include all the information required to plot the graphs as indicated in the attached main and supplementary figures. 

Mass photometry

Borosilicate microscope glass plate (No. 1.5 H thickness, 24 × 50 mm, VWR) were cleaned by sequential soaking in Milli-Q-water, isopropanol, and Milli-Q-water followed by drying under a stream of clean nitrogen. Silicone gaskets (CultureWell Reusable Gasket, Grace Bio-Labs) were placed on the clean coverslip to create a well for sample delivery. All measurements were performed on a TwoMP mass photometer (Refeyn Ltd). To convert optical reflection-interference contrast into a molecular mass, a known protein size marker (NativeMark Unstained Protein Standard, Invitrogen) was measured on the same day. For mass measurements, gaskets were filled with 18 μl of measurement buffer (25 mM Tris-HCl pH 7.5, 50 mM NaCl) to allow focusing the microscope onto the coverslip surface. Subsequently, proteins were added to a final concentration of 50 nM (final volume, 20 μl), and sample binding to the coverslip surface was monitored for 60 seconds using the software AcquireMP (Refeyn Ltd). Data analysis was performed using DiscoverMP software (Refeyn Ltd).

Magnetic tweezers assays

The sequences of the azido-oligonucleotides and the DBCO-oligonucleotides (ATDBio) used for assembly of the DNA construct are listed in Supplementary Table 1. The assembly was done similarly to existing protocols, with the only difference being that the click reaction between the azido groups and DBCO groups was considered to be essentially completed. The click reaction products were used directly as primers during the PCR. As such, a purification step was omitted, with no detectable change in the quality of the materials prepared.

All assays were conducted at 34 °C in a filtered reaction buffer containing 20 mM HEPES-KOH pH 7.8, 100 mM KCl, 5 mM MgCl₂, 1 mM DTT, 0.05% Tween-20, and 0.5 mg/ml BSA. To prepare bead-DNA constructs, an aliquot of the DNA construct (1 µl at 300 pM) was diluted using reaction buffer to obtain a 10 µl solution. In parallel, 7 µl of magnetic beads (Dynabeads^TM MyOne^TM Streptavidin C1 at 10 mg/ml, Invitrogen) were washed with 200 µl of reaction buffer, then resuspended in 25 µl of reaction buffer and mixed with the DNA solution. The 35 µl bead+DNA solution was then injected into the functionalized reaction chamber and incubated for 10 min at room temperature. Beads that remained untethered to the surface after the incubation were washed out using 4 ml of reaction buffer. The slide was then placed under the magnetic trap.

Using the PicoJAI magnetic bead tracking software suite (PicoTwist S.A.R.L), 30-60 beads were picked and tracked over time under cycles of low (~ 0 pN) and high (4 pN) forces for 60 and 30 seconds, respectively (Supplementary Fig. 4a). The force cycle was stopped at 4 pN, and the proteins were injected as a 400 µl solution containing 1 nM Ku, 0.3 nM Dnl4-Lif1 variants, 50 nM Nej1 and 1 nM MRX. After 5 min, when flow through the reaction chamber has ceased and the magnetic beads have recovered their equilibrium positions, the force cycle was relaunched, and events were measured.

The data were analyzed using a homemade Python script that detects events using the derivatives of the measured points. First, a blank measurement was performed using a bead stuck to the surface that was picked in the experiment. Then, manually, each event detected by the script can be visualized, added, erased, or modified if necessary. Three kinds of events can be considered: the first category groups events that start and end during the high-force step; the second category includes events that start during the high-force step and end during a later, distinct high-force step, with the maximum duration being 90 seconds; the third category corresponds to events in the first or second category but for which the synapsis breaks during the low-force portion of the cycle.  In the latter case, the endpoint of the event was manually annotated as the midpoint of the low-force phase. An event was characterized by its duration (in seconds) and its amplitude (in nanometers). With the 670 bp bridge used for this construct, the distribution of amplitudes for specific events followed a Gaussian with a mean of ~180 nm (for a 4 pN extending force) and a standard deviation of about 10 nm. Only events displaying an amplitude between 130 and 250 nm were retained for analysis. The distinction between specific and non-specific synaptic events was performed as previously described. Moreover, beads were analyzed only if they were tethered by a DNA molecule displaying a contour length between 0.7 and 1.3 µm (as measured at 4 pN extending force). Analysis of event distributions, survival curves, plots of DNA length were then performed using Excel or Python.

Nanofluidic assays

Reactions were carried out in 10 µl volume of 2X reaction buffer containing 50 mM Tris-HCl pH 7.74, 4 mM MgCl₂, 2 mM DTT, 0.5 mg/ml BSA, and 2 mM ATP, and 0.2 nM substrate (in molecules, using 20 kbp-long dsDNA with 4 nt-long overhangs). Lower MgCl₂ concentrations are used in this assay, as Mg²⁺ ions can interfere with the fluorescent dye YOYO-1. Ku was first preincubated with DNA for 3 min on ice. Afterwards, the rest of the proteins were added, and the reactions were incubated for 1 hour at 25 °C. For the control (DNA only) experiment, the proteins were replaced with an equivalent volume of 1X reaction buffer. For the experiments where MRX was not added, an equivalent volume of 1X reaction buffer was used. After protein incubation, DNA was fluorescently labelled by incubating with 20 µM YOYO-1 dye at 1:5 dye: base-pair ratio for 30 min at 50 °C. Finally, 0.05X TBE buffer and 2% β-mercaptoethanol were added to make a 50 µl loading solution.

Experiments were performed using nanofluidic devices that consisted of two microfluidic channels, each connecting two loading reservoirs. 200 nanofluidic channels span the microfluidic channels. Each nanochannel is 150 nm in width, 100 nm in depth and 500 µm in length, and the spacing between two consecutive nanofluidic channels is 800 nm. Images were collected using an inverted fluorescence microscope (Axio Observer Z1, Carl Zeiss AG) illuminated with a Colibri 7 LED light source (Carl Zeiss AG) with 512 x 512 pixels, 16 µm² EMCCD camera (Photometrics Evolve, Teledyne Photometrics). A FITC filter (470/20 nm excitation and 530/50 nm emission) was used. 20 frames (100 ms each) were collected for each DNA molecule. Image analysis and plotting were performed using home-written MATLAB code. Briefly, the images obtained were converted from .CZI to .TIFF file format, followed by generation of kymographs from the 20-frame images of individual DNA molecules. Kymographs were checked for quality, and DNA molecules fragmenting during image acquisition were discarded. Remaining kymographs from intact DNA molecules were used to obtain the plots shown in Supplementary Fig. 4b (ii-iv), where DNA length (µm) is on the X-axis, and the intensity (arbitrary units) on the Y-axis.

For each experimental condition, two replicate datasets are presented where one dataset is represented by light gray filled circles and the other by black unfilled circles. Each dataset comprises 250 to 300 data points from single DNA molecules. In the DNA-only control experiment, most of the molecules are ~3 µm in length and clustered around an intensity per pixel of ~160 to 190 units. The shorter-length molecules are likely from DNA fragmented during sample preparation and handling, and the lower-intensity molecules are likely from partially stained DNA. Both populations are small in comparison and are commonly observed in single-molecule nanofluidic experiments.