The C-terminus of PprI was labeled with the photoconvertible fluorescent protein mMaple3 (a gift from Stephanie Weber Lab) using homologous recombination (Manko et al., 2020; Figure 1 and Supplementary Table 1). Four PCR fragments were ligated by ClonExpress^® Ultra One Step Cloning Kit (Vazyme Biotech Co., Ltd., Nanjing), and then chemically transformed into D. radiodurans competent cells and selected by 20 μg/mL kanamycin (Sangon Biotech, Shanghai) for positive insertions. Based on multi-genomic D. radiodurans (Cox and Battista, 2005), the transformed colonies were further screened twice consecutively with selective medium containing 30 and 40 μg/mL kanamycin. The successful clones were confirmed by sequencing that mMaple3 and kanamycin resistance fragments have been integrated into the genome. The same method was used to construct a D. radiodurans strain in which the PprI protein was tagged with PAmCherry in situ.

For E. coli, the sequence of mMaple3 was inserted into the pET-28a vector using restriction enzyme (EcoRI/HindIII) (a gift from Jiang Shijie’s lab at Southwest University of Science and Technology) and then transformed into DH5α competent cells (Shanghai Weidi Biotechnology Co., Ltd, Shanghai) by heat shock method (Froger and Hall, 2007). Finally, the successful reconstituted vector was transformed into BL21 (DE3, Shanghai Weidi Biotechnology Co., Ltd, Shanghai) competent cells for mMaple3 expression.

Deinococcus radiodurans strains were streaked onto TGY plates (10 g/L tryptone, 1 g/L glucose, 5 g/L yeast extract, 15 g/L ager) with kanamycin (20–40 μg/mL). Individual colonies were inoculated into Tryptone Glucose Yeast Agar Medium (TGY) or minimal medium (MM) (Venkateswaran et al., 2000; Supplementary Table 2) and grown at 30°C for 24 h, then diluted 100-fold into new TGY or MM medium and grown to the prophase of logarithmic growth (pro-log, at OD₆₀₀ 0.3–0.5). Glass coverslips (VWR^® Micro Cover Glasses, No. 1 Slides, 0.17 mm) for D. radiodurans imaging were calcined in a muffle furnace at 500°C for 30 min in order to remove any background particles that were fluorescent in nature. A total of 1% agarose pads were prepared by mixing 2% low fluorescence agarose (Bio-Rad, Berkeley, California) with 2 × of either TGY, phosphate buffer (PBS), or MM medium (Zawadzki et al., 2015). For bacteria stained with SYBR Gold, 1 μL of 1:100 prediluted SYBR Gold was pipetted into 999 μL of MM liquid medium to make 1:10⁵ SYBR Gold culturing medium, in which the centrifuged bacteria were resuspended. The bacteria imaged on the PBS pad were washed six times with TGY-grown cells in PBS solution (Floc’h et al., 2018). Agarose pads between two glass coverslips were used to immobilize the cells during imaging processes. For BL21 individual colonies were picked and plated in 2 mL of M9 glycerol (0.2%), cultured overnight at 37°C until the OD₆₀₀ was 0.4–0.6, and then diluted in fresh M9 to grow to an OD₆₀₀ of 0.1. For fixation, the cells were centrifuged and resuspended in 4% paraformaldehyde in PBS buffer (Aladdin, Shanghai) and shaken at 22°C for 45 min. Cells fixed in this way were washed with PBS and immobilized on agarose pads (Uphoff et al., 2013; Zawadzki et al., 2015; Stracy et al., 2016).

Total internal reflection fluorescence (TIRF) microscopy is widely used in single-molecule imaging because it reduces the inevitable defocusing fluorescence problem in fluorescence imaging systems (Stracy et al., 2014). In TIRF illumination, the excitation beam is reflected at the interface between the coverslip and the imaging medium so that the resulting evanescent excitation extends only 150 nm into the sample (Paige et al., 2001). To image target proteins in bacteria, we employed the near-TIRF system, also called highly inclined and laminated optical sheet (HILO) microscopy, which provides thin excitation light at a shallow angle to the coverslip and illuminates a few microns into the sample (Tokunaga et al., 2008). The custom-built imaging system was implemented with a 405 nm laser (MDL-III-405 100 mW, CNI Laser Co. Ltd) for photoactivation of mMaple3-fused proteins and a 561 nm laser (OBIS 561 nm LS 80 mW, Coherent) for excitation on an inverted microscope platform (IX83, Olympus). The photoactivation rate was controlled by a low level of 405 nm laser at a power of 0.02–0.08 W/cm². Typical excitation power was 2.7 mW with the 561 nm laser at the sample, corresponding to a power density of 0.075 kW/cm² at the sample, which allowed adequate signal and minimal photobleaching of mMaple3. While a high excitation power of 23.0 mW was used to remove the pre-existing fluorescence molecules prior to the photoactivation. A total of 405 nm and 561 nm lasers were collinearly combined and focused at the back focal plane of an oil-immersion objective lens (UPLAPO100XO, Olympus) through a dichroic mirror (ZT405/488/561rpc, Chroma). Fluorescence emission was filtered by a multiple band filter (ZET405/488/561x, Chroma) and an additional long-pass filter (ET575lp, Chroma) in front of an EMCCD (IXON-L-897, 512 × 512 pixels; Andor). An LED light source and a spotter (M660L4-C1, Thorlabs) were implemented for bright-field images. All the images were acquired by the EMCCD camera with a pixel size of 160 nm after a 100X objective (UPLAPO100XO, Olympus). Sample position and focus were controlled by a motorized piezoelectric stage, z-motor objective mount, and autofocus system (MS-2000, PZ-2000FT, CRISP, ASI Imaging). For single-molecule tracking experiments, an exposure time of 15.7 ms/frame was used throughout this article (unless otherwise specified).

Prior to the acquisition of single-molecule data, the pre-existing autofluorescence signal as well as mMaple3 fluorescent signal and autoblinking molecules in D. radiodurans cell were pre-bleached using 561 nm laser with high excitation power (23.0 mW). mMaple3 proteins were then repeatedly activated using a continuous low 405 nm and excited with a low 561 nm activation power (2.7 mW) to ensure that the single-molecule data was collected from freshly activated mMaple3 proteins by the 405 nm laser. Over 30,000 frames were typically recorded until no new activations were observed to acquire the vast majority of PprI molecules. For SYBR Gold-stained DR-PprI-mMaple3 bacteria, the images of nucleoid were acquired by briefly turning on 0.8 mW of 488 nm laser after locating the cells under bright field. The PprI-mMaple3 molecules were then imaged again as described above, so that co-localization of the nucleoid with PprI-mMaple3 molecules could be analyzed at single-molecule level. For the DR-PprI-PAmCherry strain, the same imaging conditions as for the DR-PprI-mMaple3 strain were used, to quantitatively compare the characteristics of mMaple3 and PAmCherry.

To analyze the background values, 5 background regions of 4 × 4 pixels without cells were selected randomly from the whole field of view of 256 × 256 pixels in the TGY, PBS, and MM environments for both wild-type (DR-WT, without introducing any extrinsic fluorescent markers) and fluorescently labeled D. radiodurans (DR-PprI-mMaple3), respectively (Supplementary Figure 1). The Measure Stack function in ImageJ was used to obtain the average intensity values of the intercepted 2000 frames (1 frame equals to approximately 0.015 s, total 30 s), and the Plot function was used in Matlab to draw the time traces of the background intensity changing with time (Figures 2A, B). Four regions of 50 × 50 pixels were selected from one specific frame for each imaging condition (Supplementary Figure 2). Background distributions in four regions were separately analyzed in each case (see Supplementary Figure 4). Then, the background values of the four regions were pooled together to obtain the histograms (Figures 2C, D). These six datasets were split into two groups based on strain type, and histograms of corresponding background intensities as a function of pixels number were plotted using Matlab (Figures 2C, D).

To analyze the intensity time traces of autoblinking and mMaple3 molecules, 4 × 4 pixels regions with fluorescence signal (Supplementary Figure 3) were cropped from 2000 frames (1 frame equals to approximately 0.015 s, total 30 s) of 256 × 256 pixels in TGY, PBS, and MM environments for both DR-WT and DR-PprI-mMaple3, respectively. The time traces of the fluorescence intensity were plotted using the Plot function in Matlab (Figures 3A, B). To analyze the fluorescence intensity of autoblinking and PprI-mMaple3 molecules under different imaging environments, over 20 fluorescent signal molecules were selected in the above original video, and the Duplicate function in ImageJ was used to intercept the fluorescence signal frames from activation to irreversible bleaching (4 × 4 pixels). Use the Measure Stack function to extract the maximum signal strength value of each frame. The histogram function of Matlab was used to draw the corresponding fluorescence signal intensity normal distribution histogram (Figures 3E, F). Both Figures 2, 3A, B, E, F were analyzed from the same batch of experimental data with identical imaging conditions.

In the video of DR-PprI-mMaple3 bacteria acquired under MM imaging environment, 40 regions of 4 × 4 pixels were selected at the cell membrane or cell septum, and the average intensity of each 4 × 4 pixels was extracted using ImageJ. For autoblinking, the time traces of intensity were used to determine the background values. The number of duration frames whose average intensity is greater than the background value is calculated by Matlab and one frame is allowed to be lost. The hist function is used to obtain the histogram of the duration frame of each autoblinking molecule (Figure 3C). The histogram of the number of duration frames of the PprI-mMaple3 molecule was obtained in the same way by selecting 100 regions of 4 × 4 pixels of the fluorescence signal in the cell (Figure 3D).

Combining the single-molecule tracking technology with the PALM-centered photoactivation strategy, densely localized molecules can be imaged and tracked (Manley et al., 2008). Single-molecules are sparsely photoactivated and imaged for multiple frames until they are irreversibly photobleaching. Based on the blinking properties of mMaple3 molecules, we used a one-frame memory parameter to account for the transient disappearance of fluorophore images within the trajectory due to blinking or missed localizations (Uphoff et al., 2014; Stracy et al., 2015). Trajectories were created by connecting the molecular localizations appearing on consecutive frames of images. When other molecules are randomly activated, the process is repeated until all the molecules of interest are imaged. Using a lower excitation intensity allows the molecules to be tracked for a longer duration (Stracy et al., 2014).

Single-molecule tracking analysis was carried out using ImageJ and custom-written Matlab software (Crocker and Grier, 1996). The outline and midline of the cells were obtained by segmenting the cells with edge detection using Microbe Tracker software on bright field images (Sliusarenko et al., 2011), and the x-axis was defined as the short axis of the cells and the y-axis as the long axis of the cells to determine the position of the molecular trajectories relative to the midline (Sliusarenko et al., 2011). Molecular trajectories were plotted using Matlab, and the diffusion rates of molecules were color-coded. We followed the molecules only within the boundaries of cells and identified the fluorophore images for localization through band-pass filtering and applying intensity thresholds to each frame of the movie. If the cell contains bright spots that are not bleached during 561 nm pre-bleaching, or if the cell has an abnormally high fluorescence background, it is excluded from the analysis. PprI-mMaple3 molecules and autoblinking molecules were detected, fit to symmetrical 2D Gaussian point spread functions (PSFs), and tracked using Matlab. A custom script was applied to create regions of interest (ROIs) based on the cell masks generated by Microbe Tracker. All of the localizations within each ROI were then tracked and analyzed separately. Tracks were linked to trajectories if they appeared in successive frames in a 3-pixels (0.48 μm) window. Moreover, we used a one-frame memory parameter to minimize the transient loss of fluorophore images within the trajectory due to blinking or loss of positioning.

We distinguished immobile and diffusing proteins by calculating an apparent diffusion coefficient D* = MSD/(4Δt) from the mean-squared displacement (MSD) and taking only trajectories with at least 4 steps and Δt = 15.7 ms, Δt is the lag times. For each measurement, the σ_loc of localization precision is expressed as a positive offset (Michalet and Berglund, 2012; Garza de Leon et al., 2017) in the D* value of the σ_loc²/Δt. We used the same D* value of immobile molecules to constrain one species and fit the data well for a second and/or third unconstrained species (Supplementary Figure 5).

In order to obtain the localization error, 100 mMaple3 molecules were measured in the fixed E. coli cells expressing the plasmid carrying mMaple3. Each molecule presented a cluster of localizations owing to continuous acquisition of the same molecule. Localizations from 100 clusters (each containing >8 localizations) were aligned by their center of mass to generate the 2D presentation of the localization distribution. Histograms of distribution in x and y were fit to Gaussian functions, and the resultant s.d. (σx and σy) was shown. This data evaluated the single-molecule localization accuracy (Supplementary Figure 6) and was used to distinguish between immobile and diffusing PprI-mMaple3 molecules.

As mentioned elsewhere (Vrljic et al., 2002; Stracy et al., 2015, 2016; Garza de Leon et al., 2017), we obtained the diffusion coefficient D₁* by fitting the probability density of the sample D* distribution with the analytical equation of D₁*. The analytical expression of the single-mode case is:

where x is the empirical D* data. It is hypothesized that there may be three species of PprI with different rates of movement: molecules that diffuse rapidly into the cytoplasm, slowly moving DNA-searching molecules, and PprI that binds specifically to DNA. As a result, we introduced a third.

where D₁*, D₂*, and D₃* are the diffusion coefficients for the three different species, A₁, A₂, and A₃ are the fractions of the molecules found in each of the states, and A₁+A₂+A₃ = 1, respectively.

Culturing media greatly influences the growth status of bacteria, and it has been shown that the rich growth media could contribute to autoblinking (Floc’h et al., 2018). Although photoconvertible fluorescent protein has been widely applied in single-molecule tracking with good photophysical characteristics and optimal cell compatibility, PAmCherry had been identified as a non-ideal fluorescent probe for D. radiodurans in previous studies (Liao et al., 2015; Garza de Leon et al., 2017; Floc’h et al., 2018). We employed another photoconvertible fluorescent protein, mMaple3 with better photostability and brightness (Wang et al., 2014), to label PprI protein (DR-PprI-mMaple3) in D. radiodurans (Figure 1). To verify the effect of imaging media on the detection of fluorescent proteins, we measured background intensity in wild-type D. radiodurans (DR-WT) and DR-PprI-mMaple3 cells under different imaging media using TIRF microscopy (Figures 2A, B). In TGY, the background intensities stand at the highest level for both DR-WT and DR-PprI-mMaple3, with the average of 234 ± 5.03 and 235 ± 5.11 a. u., respectively (Figures 2A, B). In comparison, the background intensity in PBS or MM were much lower, ∼150 a. u., which was dramatically different from that in TGY. For DR-WT, the background intensity of 141 ± 2.77 a. u. in PBS was slightly lower than that of 154 ± 3.25 a. u. in MM (Figure 2A); whereas in DR-PprI-mMaple3, the background intensity of 157 ± 3.30 a. u. in PBS was similar as that of 148 ± 3.15 a. u. in MM imaging medium (Figure 2B). The background intensities in PBS and MM showed no significant differences, which was independent of imaged strains, suggesting both washed cells and growing cells in MM, instead of TGY, were effective for reducing the background. In addition, the time traces of background intensities showed no visible drift, indicating a constant background with a reliable imaging system.

To verify whether the averaged background intensity of the entire field of view is independent of spatial locations, we randomly select 4 regions of 50 ± 50 pixels in one frame (Supplementary Figure 2). It can be found that the background levels are similar in different regions, demonstrating the uniformity of the spatial distribution of background (Supplementary Figure 4). Histograms of the spatial distribution (Figure 2) showed that the mean background intensity from DR-WT samples under TGY, PBS, and MM were 251 ± 26.78, 139 ± 12.17, and 153 ± 13.71 a. u. respectively (Figure 2C). And the mean background intensity of DR-PprI-mMaple3 in TGY, PBS, and MM were 235 ± 22.37, 155 ± 14.70, and 146 ± 13.37 a. u. (Figure 2D). These results suggested that the mean background intensity of spatial distribution was consistent with that of temporal distribution (Figures 2A, B). Furthermore, both strains growing in PBS and MM showed greatly reduced background, allowing a good imaging condition for acquiring single-molecule signals from D. radiodurans.

To investigate the brightness level and fluorescence stability of autoblinking molecules and PprI-mMaple3 molecules, we observed the fluorescence intensity trend of the two species from DR-WT and DR-PprI-mMaple3 samples in TGY, PBS and MM imaging medium for 2000 frames in the selected fluorescence region (Figures 3A, B). And we compared the fluorescence intensities of the two species in three imaging environments (Figures 3E, F, Supplementary Figure 3). Among them, the fluorescence signals in DR-WT bacteria were all autoblinking molecules, while the fluorescence signals in DR-PprI-mMaple3 bacteria might have been mixed with a small amount of autoblinking molecules. In DR-WT samples, the single-molecule autoblinking intensity was up to 572 ± 70.40 a. u. in TGY imaging medium, while it decreased to 372 ± 61.70 a. u. and 382 ± 45.02 a. u. in PBS and MM (Figure 3E), respectively (Supplementary Table 3). We further quantified the signal-to-background ratios (SBR), defined as the ratio of the peak intensity to the averaged baseline background of the same acquired region (Figure 3B). For DR-WT bacteria, the SBRs of autoblinking molecules in TGY, PBS, and MM were 2.00, 1.98, and 2.05 (Supplementary Table 3), respectively, indicating that the SBRs did not show much difference among 2000 selected frames from DR-WT cells (Figure 3A). In contrast, there was little difference in intensity of PprI-mMaple3 molecules in TGY, PBS, and MM, which were 664 ± 160.45 a. u., 606 ± 117.32 a. u., 661 ± 117.87 a. u. (Figure 3F and Supplementary Table 3), respectively, indicating that the fluorescence intensity of mMaple3 was unaffected by the bacterial growing conditions. However, SBRs of PprI-mMaple3 increased in PBS and MM imaging medium due to the reduced background noise (Figure 3B), which were 2.96, 3.51, and 3.58 (Supplementary Table 3), respectively. These results indicated that the mMaple3 molecules in PBS and MM imaging medium were brighter than autoblinking molecules, and the higher SBR made the fluorescence signal of PprI-mMaple3 more visible in PBS and MM imaging medium. However, PBS imaging medium with zero nutrition might affect D. radiodurans growth, leading to abnormal expression of PprI. Therefore, MM medium was considered as the ideal imaging medium for our single-molecule tracking experiments in D. radiodurans cells, and subsequent experiments in this paper were imaged in MM medium.

The SBRs and the number of localizations (Supplementary Table 4) at the same number of acquisition frames (14,654 frames) were compared between the PprI-mMaple3 molecules and the PprI-PAmCherry molecules using the same method. The result shows that mMaple3 provided a much larger number of PprI localizations, which was ∼4-fold higher than that of the PprI-PAmCherry molecules in the same imaging period. Comparing the SBRs of mMaple3, PAmCherry, and autoblinking, it is clear that the SBR of autoblinking is only slightly lower than that of PAmCherry, making these two hardly distinguishable.

To further evaluate the photostability of these two species, we calculated the duration frames of autoblinking (Figure 3C) and PprI-mMaple3 molecules (Figure 3D), respectively. The mean of duration frames was only 1.41 for autoblinking molecules (Figure 3C), while the average duration frames of PprI-mMaple3 molecules were 8, with the maximum reaching 54 frames (Figure 3D), showing a huge difference in duration frames between autoblinking and PprI-mMaple3 molecules. Consequently, the majority of autoblinking molecules (trajectories ∼4 steps) could be removed based on the duration frames of the fluorescent molecules (autoblinking molecules and mMaple3 molecules) during the data analysis.

Although the autoblinking molecules have lower signal intensity and shorter average duration frames compared to the PprI-mMaple3 molecules, these two parameters can remove the majority of the autoblinking molecules, and a tiny fraction of those ones with stronger signal intensity and longer duration frames still remain. In order to quantify the influence of residual autoblinking molecules, we analyzed the spatial distributions and diffusion coefficients of autoblinking and PprI-mMaple3 molecules. PprI-mMaple3 molecules tend to distribute in various regions of cells without clear bias, whereas autoblinking molecules localize preferentially on cell membrane. In addition, the total number of PprI-mMaple3 molecules was dominantly larger than autoblinking molecules acquired during the same acquisition time (Figure 4A), suggesting autoblinking molecules only contribute <6.5% molecule localizations to the recorded signals. Most of the autoblinking molecules can be removed from the collected signal by the above-mentioned filtering process of intensity and duration frames.

In living cells, PprI molecules may exist in different motion states, depending on their specific functionalities. Based on previous studies, it is known that most proteins are in two states of motion. Therefore, we first classified PprI molecules into two species and free-fitted to obtain a diffuse population with D* > 0.11 μm²/s and an immobile population with D* < 0.11 μm²/s (Stracy et al., 2021). In 32766 frames, we collected 37942 PprI-mMaple3 molecules in total, making up 2532 trajectories, of which diffusing molecules accounted for 89.8% of the total. A total of 1935 autoblinking molecules were observed, constituting 165 single-molecule trajectories, which equals 6.5% of that from PprI-mMaple3 (Figures 4B–D). Therefore, autoblinking signals cover only a tiny fraction of overall mMaple3 molecules, ∼5.1% for localizations and ∼6.5% for linked trajectories. For the categorized diffusing species, only 80 molecules were linked as trajectories, which equals ∼3.5% of that formed within diffusing PprI-mMaple3 molecules (Figure 4D), which is largely negligible for single-molecule imaging. Interestingly, we observed that more autoblinking molecules (1098/1935, ∼56.7%) stay as immobile state, while PprI-mMaple3 molecules tend largely to be the diffusing state (34066/37942, ∼89.8%). The trajectory lengths of PprI-mMaple3 molecules were significantly longer than those of autoblinking molecules, which can serve as another criterion for separating mMaple3 from autoblinking signals. The trajectory lengths of immobile molecules were slightly longer than those of diffusing molecules in both autoblinking and PprI-mMaple3 molecules (Figure 4E).

We stained the nucleoid of D. radiodurans cells using SYBR Gold and acquired nucleoid fluorescence images under 488 nm laser irradiation (Figure 5A, green). Overlaying (Figure 5D) the molecular trajectories of Pprl-mMaple3 (Figure 5B) with the position of the nucleoid grayscale maps (Figure 5C), Matlab-based analysis yielded that the majority of Pprl molecules preferentially covered the nucleoid region, with diffusing molecules being more pronounced. In order to quantitatively analyze the relative positions of PprI proteins with respect to the nucleoid in D. radiodurans cells in the normal state, the PprI proteins were categorized into two types: immobilized molecules and diffusing molecules, according to the apparent diffusion coefficient, and the numbers of molecules inside and outside the nucleoid for these two motion states were calculated, respectively (Figure 6). It was found out that 65.6% of the diffusing molecules localized within the nucleoid, whereas the tendency of the immobilized molecules was not significant (Figure 6B). Taken together, a larger fraction of PprI molecules co-localized with the nucleoid, accounting for ∼62.1% of the total number of PprI molecules (Figure 6C).

To measure the mobility of PprI molecules in living D. radiodurans, we performed single-molecule tracking with DR-PprI-mMaple3 strain in MM (Figure 7A). We first calibrated the localization error of our microscope setup by measuring the mMaple3 molecules in fixed E. coli cells and obtained the averaged D* was ∼0.06 μm²/s and the localization error as ∼30 nm (Supplementary Figure 6). Instead of an inappropriate two-species fitting (Supplementary Figure 5), the D* distribution of PprI-mMaple3 molecules was fitted into a three-species model (Figure 7B) with R² = 0.9245 (Figure 7C), shown as the immobile molecules (18.6%, D* = 0.07 μm²/s), slow diffusing molecules (42.5%, D* = 0.20 μm²/s), and fast diffusing molecules (38.9%, D* = 0.72 μm²/s) (Figure 7B). The residual plot shows the fitting is optimal (Figure 7C). From the color-coded trajectories of PprI molecules, we found that the majority (>80%) of PprI molecules belong to the fast and slow diffusing species, which tend to search freely for the target in the nucleoid-free regions and to bind transiently the target of DNA or protein partners within the nucleoid regions (Figure 7D). However, the immobile species is supposed to engage into active reaction with DNA or protein partners in the DNA repair pathway (Figure 7D).