Visualizing Protein-DNA Interactions in Live Bacterial Cells Using Photoactivated Single-molecule Tracking

Protein-DNA interactions are at the heart of many fundamental cellular processes. For example, DNA replication, transcription, repair, and chromosome organization are governed by DNA-binding proteins that recognize specific DNA structures or sequences. In vitro experiments have helped to generate detailed models for the function of many types of DNA-binding proteins, yet, the exact mechanisms of these processes and their organization in the complex environment of the living cell remain far less understood. We recently introduced a method for quantifying DNA-repair activities in live Escherichia coli cells using Photoactivated Localization Microscopy (PALM) combined with single-molecule tracking. Our general approach identifies individual DNA-binding events by the change in the mobility of a single protein upon association with the chromosome. The fraction of bound molecules provides a direct quantitative measure for the protein activity and abundance of substrates or binding sites at the single-cell level. Here, we describe the concept of the method and demonstrate sample preparation, data acquisition, and data analysis procedures.


Introduction
This protocol describes the direct measurement of protein-DNA interactions in living Escherichia coli cells. The technique utilizes the change in the diffusion coefficient of a single fluorescently labeled protein as it binds the chromosome (Figure 1). To demonstrate the method we use DNA polymerase I (Pol1), a prototypical DNA-binding protein that fills DNA gaps in lagging strand replication and excision repair pathways 1 .
The advent of super-resolution fluorescence microscopy enables visualization of molecular structures in cells with nanometer resolution. Photoactivated Localization Microscopy (PALM) employs fluorescent proteins that can be activated from an initial dark state to a fluorescent state (Figure 2). Only a subset of all labeled molecules is activated at any time to determine their positions in a sequential manner, independently of the total concentration of labeled molecules in the sample 2 . The localization precision per molecule mainly depends on the size of the fluorescent Point Spread Function (PSF), the number of collected photons, and the background signal 3 . Many applications of this method focus on the improved visualization of cellular structures. The realization that PALM can be combined with single-molecule tracking 4 opened new avenues to directly follow the movement of arbitrary numbers of labeled proteins in living cells. Increased sensitivity and temporal resolution of fluorescence microscopes now allow tracking of single diffusing fluorescent proteins in the bacterial cytoplasm 5 .
Here, we employ PAmCherry, an engineered fluorescent protein that irreversibly converts from an initial nonfluorescent state to a fluorescent state upon irradiation with 405 nm light 6 . Activated PAmCherry fluorophores can be imaged by excitation at 561 nm and tracked for several frames until photobleaching. We demonstrate the ability of the method to identify transient DNA-binding events of single proteins using a fusion of Pol1 and PAmCherry. Treatment of cells with methyl methanesulfonate (MMS) causes DNA methylation damage that is turned into gapped DNA substrates by base-excision repair enzymes. Our method shows clear binding of single Pol1 molecules in response to MMS damage 7 .

Cell Culture
Use sterile culture tubes and pipette tips. The E. coli strain AB1157 polA-PAmCherry carries a C-terminal PAmCherry fusion of Pol1. The fusion was inserted at the native chromosomal location by replacing the wild-type gene using lambda-Red recombination as described in Datsenko et al. 8 Functionality of the fusion protein was confirmed as judged by cellular growth rates and sensitivity to the DNA damaging agent methyl 1. Find a new field of view (FOV) of cells in transmitted light microscopy mode and focus the image. Take a camera snapshot to record the cell outlines ( Figure 4A). 2. Cover the sample from ambient light and switch on the EMCCD camera gain. 3. Switch on the 561 nm laser and bleach the cellular autofluorescence and background spots on the coverslip for a few seconds before starting data acquisition. For cells grown and imaged in M9 medium and using burned coverslips there is usually very little fluorescence background; however, prebleaching could be useful for imaging cells in a rich growth medium such as LB. Note that intense illumination is toxic to cells so prebleaching should be kept to a minimum. 4. Start the acquisition of a PALM movie under continuous 561 nm excitation at 15.26 msec/frame. 5. Switch on the 405 nm laser and gradually increase the intensity over the course of the movie, reaching up to 1 W/cm 2 . Avoid higher 405 nm intensities that cause cellular autofluorescence. Pay attention to the density of fluorescent molecules -it is important to keep activation rates low such that PSFs are clearly isolated in each frame (Figures 4D-F). 6. Record 10,000 frames/movie (depending on the number of molecules to be imaged per cell); one movie typically takes 2-3 min and requires 0.5-1 GB of hard disk space depending on the size of the FOV. 7. Repeat the acquisition procedure for multiple FOV. Note that each FOV can only be imaged once because PAmCherry fluorophores get photoactivated and bleached irreversibly.

Data Analysis
An automated and robust data analysis framework is essential for the performance and efficiency of the method. We use custom software written in MATLAB.
1. Perform the localization analysis using algorithms described in Crocker et al. 12 , Holden et al. 13 , HoldenI et al. 14 , and Wieser et al. 15 PSFs are first identified in a band-pass filtered image using a Gaussian kernel with 7 pixels diameter ( Figure 5A). Candidate positions correspond to PSFs with peak pixel intensities above 4.5 times the standard deviation of the background signal ( Figure 5B). The locally brightest pixel per candidate PSF serves as initial guess for fitting an elliptical Gaussian function ( Figure 5C). The free fit parameters are: x-position, y-position, x-width, y-width, rotation angle, amplitude, and background offset. The elliptical Gaussian mask accounts for molecule during the exposure time, which blurs and deforms the PSF. 2. Plot the resulting (x, y) localizations from all frames of the PALM movie onto the transmitted light microscopy image of the same FOV.
Localizations of Pol1-PAmCherry should appear within the central area of E. coli cells (Figure 6A). If many localizations appear outside of cells, the localization threshold was set too low or the sample contained background fluorescent particles. 3. For automated tracking analysis, the MATLAB implementation of the algorithm described in Crocker et al. 12 can be used (see "Diffusion analysis" in the Discussion section). Positions that appear in subsequent frames within a user-defined tracking window are connected to form a trajectory. In the case that multiple localizations occur in the same window, tracks are uniquely assigned by minimizing the sum of step lengths. For a detailed discussion of the various considerations when calculating diffusion coefficients from single-particle tracking data, see Wieser et al. 15 1. The algorithm uses a memory parameter to account for transient blinking or missed localizations during a track. Here, we set the memory parameter to 1 frame; higher values can be used for tracking fluorophores with long-lived dark states. 2. Choose a suitable tracking window based on the following calibration steps. For Pol1, we use 0.57 µm (5 pixels). 3. Run the tracking algorithm for a range of tracking window parameters. Calculate the number of measured tracks per cell as a function of the tracking window to identify the smallest possible tracking window that does not split tracks ( Figure 6B). 4. Plot the resulting tracks on the transmitted light microscopy image of the same FOV to visualize the spatial distribution of molecule movement within cells. Pol1 tracks should display diffusion confined within single cells (Figures 6C-D). 5. If a fraction of tracks appears to cross between cells this suggests that separate molecules were erroneously linked because the tracking window was chosen too large and/or the photoactivation rate was too high (Figure 6E). 6. Plot the cumulative distribution of the step lengths between consecutive localizations (Figure 6F). The curve rises and saturates smoothly for sufficiently large tracking windows but shows a cutoff edge if the window was chosen too small.

4.
To analyze the diffusion characteristics of Pol1, compute the mean-squared displacement (MSD) between consecutive localizations for each track with a total of N steps): Include only tracks with at least 4 steps (N ≥ 5 localizations) to reduce the statistical uncertainty in the MSD values. 5. Plot a curve of MSD values over a range of lag times by calculating displacements over multiple frames (Figure 6G). The shape of the MSD curve can help to classify the observed molecular motion (Figure 6H).

Representative Results
The concept of photoactivated single-molecule tracking to study protein-DNA interactions in vivo is illustrated in Figure 1. PAmCherry fusion proteins are detected in live E. coli cells in a sequential manner by photoactivating single molecules stochastically with 405 nm light at a frequency of less than one molecule per cell at a time. Activated molecules are imaged under continuous 561 nm excitation. Molecular movement in the cell can be tracked by connecting nearby localizations in a series of frames until irreversible photobleaching. Because the diffusion of DNA-binding proteins is slowed upon binding the chromosome, the apparent diffusion coefficient D* obtained per track directly reports on individual protein-DNA interactions.  Figure 4. Note that the density is not solely determined by the 405 nm intensity but additionally by the number of molecules that are available for activation; the pool of remaining molecules is depleted over the course of a PALM movie.
Localization analysis is performed for each frame of a PALM movie as illustrated in Figure 5. We measured the localization precision using immobile molecules in fixed cells or bound molecules in live cells. Our acquisition settings gave a localization precision of σ loc = 40 nm, in agreement with the theoretical prediction 3 .
The resulting Pol1 localizations occupy the central area of the cell (Figure 6A), broadly recapitulating the spatial organization of the E. coli nucleoid 7 . The majority of Pol1 tracks in undamaged cells display diffusion as shown in Figure 6C. A typical cell contains several hundred Pol1 tracks (Figure 6D), consistent with the copy number of approximately 400 Pol1 molecules per E. coli cell 1 . Figures 6B and 6E-F provide guidance on choosing a suitable tracking window parameter -if the tracking window is too large, different molecules are more likely to become erroneously linked to a track; if the tracking window is too small, tracks with longer steps will be split. The MSD curve for Pol1 rises linearly for short lag times and saturates at longer lag times due to cell confinement ( Figure 6G). Different types of molecular motion can be identified by MSD analysis. Directed motion gives a parabolic curve; Brownian motion is characterized by a straight line; the confined diffusion curve reaches a plateau; an offset of the MSD curve for immobile particles represents the localization uncertainty ( Figure 6H). Additional information on singleparticle tracking and troubleshooting tips can be found in Arnauld et al. 16 We previously applied the method to measure the DNA-repair activity of Pol1 in response to exogenous DNA alkylation damage 7 . The D* histogram of Pol1 tracks in undamaged cells shows a dominant population of diffusing molecules (Figures 7A-C). A small fraction of 2.7% bound Pol1 molecules is likely involved in lagging strand replication and repair of endogenous DNA damage. Under continuous 100 mM MMS damage, the population of tracks with D* ~ 0 µm