ATP binding facilitates target search of SWR1 chromatin remodeler by promoting one-dimensional diffusion on DNA
Cite this dataset
Carcamo, Claudia et al. (2022). ATP binding facilitates target search of SWR1 chromatin remodeler by promoting one-dimensional diffusion on DNA [Dataset]. Dryad. https://doi.org/10.5061/dryad.ghx3ffbqw
One-dimensional (1D) target search is a well-characterized phenomenon for many DNA-binding proteins but is poorly understood for chromatin remodelers. Herein, we characterize the 1D scanning properties of SWR1, a conserved yeast chromatin remodeler that performs histone exchange on +1 nucleosomes adjacent to a nucleosome-depleted region (NDR) at gene promoters. We demonstrate that SWR1 has a kinetic binding preference for DNA of NDR length as opposed to gene-body linker length DNA. Using single and dual color single-particle tracking on DNA stretched with optical tweezers, we directly observe SWR1 diffusion on DNA. We found that various factors impact SWR1 scanning, including ATP which promotes diffusion through nucleotide binding rather than ATP hydrolysis. A DNA-binding subunit, Swc2, plays an important role in the overall diffusive behavior of the complex, as the subunit in isolation retains similar, although faster, scanning properties as the whole remodeler. ATP-bound SWR1 slides until it encounters a protein roadblock, of which we tested dCas9 and nucleosomes. The median diffusion coefficient, 0.024 μm2/s, in the regime of helical sliding, would mediate rapid encounter of NDR-flanking nucleosomes at length scales found in cellular chromatin.
Dual optical tweezers and confocal microscope setup and experimental workflow
The LUMICKS cTrap (series G2) was used for optical tweezer experiments, configured with two optical traps. The confocal imaging laser lines used were 532 nm (green) and 640 nm (red) in combination with emission bandpass filters 545–620 nm (green) and 650–750 nm (red). A C1 type LUMICKS microfluidics chip was used. The microfluidics system was passivated at the start of each day of imaging as follows: 0.1% BSA was flowed at 0.4 bar pressure for 30 min, followed by a 10 min rinse with PBS at 0.4 bar pressure, followed by 0.5% Pluronic F-127 flowed at 0.4 bar pressure for 30 min, followed by 30 min rinse with PBS at 0.4 bar pressure. For SWR1 sliding on naked DNA, 4.2 µm polystyrene beads coated in streptavidin (Spherotech cat# SVP-40-5) were caught in each trap, and LUMICKS biotinylated lambda DNA was tethered. Both traps had trap stiffness of about 0.8 pN/nm. For SWR1 sliding on lambda nucleosome array, a 4.2-µm polystyrene bead coated in streptavidin was caught in trap 1, and a 2.12-µm polystyrene bead coated in anti-digoxigenin antibody (Spherotech cat# DIGP-20-2) was caught in trap 2 which is upstream in the path of buffer flow to trap 1. For this configuration, trap 1 had a trap stiffness of about 0.3 pN/nm whereas trap 2 had a trap stiffness of about 1.2 pN/nm. The presence of a single tether was confirmed by fitting a force extension plot to a worm-like chain model in real time while collecting data using LUMICKS BlueLake software. For confocal scanning, 1.8 µW of green and red laser power were used. For most traces, the frame rate for SWR1 imaging was 50 ms, whereas for Swc2 it was 20 ms. Experiments were performed at room temperature. SWR1 and Swc2 were both imaged in histone exchange reaction buffer (25 mM HEPES pH 7.6, 0.37 mM EDTA, 5% glycerol, 0.017% NP40, 70 mM KCl, 3.6 mM MgCl2, 0.1 mg/ml BSA, 1 mM BME) made in imaging buffer. dCas9 was added to the flow chamber in Cas9-binding buffer (20 mM Tris–HCl pH 8, 100 mM KCl, 5 mM MgCl2, 5% glycerol) made in imaging buffer. Imaging buffer (saturated Trolox [Millipore Sigma cat# 238813], 0.4% dextrose) is used in place of water when preparing buffers. All buffers were filter sterilized with a 0.2 μm filter prior to use.
Single-particle tracking and data analysis
LUMICKS Bluelake HDF5 data files were initially processed using the commercial Pylake Python package to extract kymograph pixel intensities along with corresponding metadata. Particle tracking was then performed in MATLAB (MathWorks). First, spatially well-separated particles were individually segmented from full-length kymographs containing multiple diffusing particles. Next, for each time step, a one-dimensional Gaussian was fit to the pixel intensities to extract the centroid position of the particle in time. Then the MSD for each time lag was calculated using:
where N is the total number of frames in the trace, n is the size of the time lag over which the MSD is calculated, i is the sliding widow over which displacement is measured, and X is the position of the particle. Since particles exhibit Brownian diffusion, the diffusion coefficient for each particle was then calculated from a linear fit to the initial portion of the MSD versus time lag plot by solving for D using: MSD=2Dt. For mean MSD plots, traces with the same frame rate were averaged together, resulting in a slightly different n value as compared to all trajectories in a condition.
For the linear fit, the number of points included varied to optimize for a maximal number of points fit with the highest Pearson correlation (r2) and a p-value lower than 0.05. For particles where this initial best fit could not be found, the first 25% of the trace was linearly fit. Fits that produced negative slope values corresponded to traces where particles are immobile; to reflect this, negative slopes were given a slope of 0. Finally, outlier traces with diffusion coefficients greater than 0.14 µm2/s for SWR1 or 5 µm2/s for Swc2 were dropped; in every case, this consisted of less than 3% of all traces. The distribution of diffusion coefficients estimated using this method was almost identical to what is produced using an alternative method which extracts diffusion coefficients using a linear fit from time lags 3–10 rejecting fits with r2 < 0.9 (Tafvizi et al., 2008) (data not shown). A summary of statistics as well as criteria for excluding traces is provided in Table 3. Also included are the number of biological and technical replicates per condition. A biological replicate is defined as a fresh aliquot of protein imaged on a different imaging day, whereas a technical replicate is the number of distinct DNAs or nucleosome arrays used per imaging condition; a single DNA could accommodate one or more fluorescently tagged proteins.
We estimated the localization precision using the following formula:
where N is the number of photons collected which was on average 12.9 photons per 5-pixel window surrounding the centroid (data not shown); s is the standard deviation of the microscope point-spread function, 294 nm; a is the pixel size, 100 nm; and b is the background intensity which was on average 0.8 photons per 5-pixel window. This results in a σ = 82 nm.
Please visit https://github.com/ccarcam1/SWR1_1D_Diffusion_Publication to download the Matlab scripts used to generate the main figures. The datasets provided are not in the raw .h5 file format since in the raw format the data is several hundreds of gigabytes, and is therefore not easily accessible. The provided datasets include raw kymograph information as well as the associated analysis of each kymograph (cropping, linear fitting, diffusion coefficients, etc.). Matlab version installed: R2021a
National Science Foundation, Award: DGE-1746891
National Institute of General Medical Sciences, Award: GM007445
NIH Office of the Director, Award: S10 OD025221
National Institute of General Medical Sciences, Award: R35 GM122569
National Institute of General Medical Sciences, Award: R01 GM125831
National Institute of General Medical Sciences, Award: T32 GM007445
National Institute of General Medical Sciences, Award: F32 GM128299
National Institute of General Medical Sciences, Award: F32 GM133151
Howard Hughes Medical Institute