Fig. 1: Live-cell imaging to study NuRD complex binding kinetics and function.

a, Schematic representation of the NuRD complex interacting with chromatin in the presence and absence of MBD3. b, Left, single JF₅₄₉-HaloTagged molecules in the NuRD complex were tracked in 3D using a double-helix point spread function microscope; two puncta are recorded for each fluorophore with their midpoint providing the lateral x, y position and the angle between them representing the axial position in z relative to the nominal focal plane (see Extended Data Fig. 2 and Supplementary Videos 1–8 for examples of the raw data). Right, examples of extracted single particle trajectories from 20 ms exposure imaging of CHD4 show periods of unconfined and confined diffusion. c, The approach used for segmentation of the single-molecule tracks; the data shown are from the 20 ms exposures of CHD4-HaloTag-JF₅₄₉. Step 1, Left, a single-molecule trajectory showing an example sliding window (blue). Right, four biophysical parameters are calculated for a sliding window that is moved through the trajectory: α, D_app, Lc and the norm∥V∥ of the mean velocity, were all estimated from a sliding window of 11 consecutive images. Step 2, Left, several trajectories with example sliding windows (blue). Right, Histograms of the values of the four biophysical parameters extracted in Step 1 from all the sliding windows computed for all the recorded trajectories. Step 3, Left, Based on the values of the four biophysical parameters (producing a four-dimensional feature space) each point in each trajectory is classified as either confined (C) or unconfined (U) using a Gaussian mixture model (GMM). The histograms from Step 2 can then be separated into confined (blue) and unconfined (orange) populations. Step 4, the posterior probability P of the GMM (Step 3) is computed on the four parameters for each sliding window Xi where the index of the trajectory is represented by i = 1,…,N (Xi(k∆t) ∈ C with P(k∆t) > 1 − P(k∆t) (blue); otherwise Xi(k∆t) ∈ U (orange)). The result is a segmented trajectory where each timepoint is assigned as confined or unconfined (see Methods for more details).