Single-step photobleaching and image analysis

15% blue laser power was used to bleach RAD-51^f clusters in minimal imaging area to obtain sufficiently high bleaching time resolution. Scans were sectioned and stacked in Fiji using a custom-written script. Maximum likelihood estimation was used to determine each of the photobleaching steps within a maximum intensity/frame n. trace as previously described (Autour et al., 2018). The step sizes were subsequently binned and the histogram was fit to a double Gaussian equation in GraphPad Prism. 7. For cluster growth analysis, individual clusters were analyzed for intensity increase in-between frames normalized to single-step intensity values. A cluster was considered as growing if the number of RAD-51^f promoters in the cluster increased by at least a single RAD-51^f protomer during the time the cluster dwelled on ssDNA. The growth frequency of RAD-51^f clusters was reported for each individual ssDNA molecule. For real-time RPA-eGFP displacement analysis, real-time force and fluorescence data were exported from Bluelake HDF5 files and analyzed using custom-written scripts in Pylake Python package. Force was down sampled to 3 Hz for plotting. For RPA-eGFP free patch edge binding analysis, custom position-analysis script was built to extract the position of individual RPA-eGFP peaks and depressions, A647 intensity peaks were then aligned and their maxima position extracted to monitor proximity to the RPA-eGFP signal depression edges. For apparent nucleation rate analysis in Figure 2E we used custom-made algorithm to quantify the Rad51 nucleation frequency from the kymograph showing eGFP-RPA displacement in time. In each kymograph, color level for the blue channel was adjusted to increase the contrast. Subsequently, the image was median-filtered in x axis (25 frames window). A negative of the image was then smoothed in y axis by the signal convolution function and subsequently the process was repeated in x axis (5 pixel window). A peak detection function was employed on the processed image to quantify the number of RAD51 filaments in time. The number of detected peaks in time was fitted with a single exponential function y = A_max (1-exp(-k^∗t)). Worm-like chain (WLC) model for λ dsDNA was used as a reference for force-extension curve comparison. Custom-written WLC fitting script was used to calculate contour length and subsequently gapped length of gapped DNA substrates. Growth rates in real-time experiments as well as dwell-times and binding frequencies were estimated in Fiji. Nucleation frequencies were plotted as a function of RAD-51^f concentration and fitted with power-law in GraphPad Prism 7. Dwell-times of RAD-51 clusters were binned into appropriate dwell-time categories cumulative survival analysis was performed using Igor 8.0. Mann-Whitney test was used to assess statistical significance of the data where appropriate.