Single-molecule motions and interactions in live cells reveal target search dynamics in mismatch repair

Significance

We integrated single-molecule superresolution imaging with biochemical and genomic approaches to understand how the mismatch repair protein MutS efficiently identifies DNA mismatches during real time in living cells. We show that MutS molecules move fast, exploring the entire nucleoid, but can transition to a slow-moving population that is localized at the replisome even before a mismatch is produced. We show that bacterial MutS must initiate mismatch binding in very close proximity to the replisome. We also show that mismatch detection increases MutS speed, supporting the model for MutS sliding clamp formation after mismatch recognition. Our results provide fundamental insight into the searching behavior of single MutS molecules during DNA replication in live cells.

Abstract

MutS is responsible for initiating the correction of DNA replication errors. To understand how MutS searches for and identifies rare base-pair mismatches, we characterized the dynamic movement of MutS and the replisome in real time using superresolution microscopy and single-molecule tracking in living cells. We report that MutS dynamics are heterogeneous in cells, with one MutS population exploring the nucleoid rapidly, while another MutS population moves to and transiently dwells at the replisome region, even in the absence of appreciable mismatch formation. Analysis of MutS motion shows that the speed of MutS is correlated with its separation distance from the replisome and that MutS motion slows when it enters the replisome region. We also show that mismatch detection increases MutS speed, supporting the model for MutS sliding clamp formation after mismatch recognition. Using variants of MutS and the replication processivity clamp to impair mismatch repair, we find that MutS dynamically moves to and from the replisome before mismatch binding to scan for errors. Furthermore, a block to DNA synthesis shows that MutS is only capable of binding mismatches near the replisome. It is well-established that MutS engages in an ATPase cycle, which is necessary for signaling downstream events. We show that a variant of MutS with a nucleotide binding defect is no longer capable of dynamic movement to and from the replisome, showing that proper nucleotide binding is critical for MutS to localize to the replisome in vivo. Our results provide mechanistic insight into the trafficking and movement of MutS in live cells as it searches for mismatches.

DNA mismatch repair (MMR) is the highly conserved process responsible for correcting DNA replication errors (1). Although replication errors occur infrequently in bacteria (∼1 error per 31 million bp) (2), the consequences of MMR failure on human health are severe (3). MutS is the first protein involved in the MMR pathway, and it is responsible for detecting rare base-pairing errors. In Bacillus subtilis, MutS then recruits MutL, an endonuclease in most bacteria and eukaryotic organisms, to incise the DNA (4, 5). After MutL incision, the error-containing strand is removed, and the DNA is resynthesized to complete error correction (6).

The mechanism by which MutS homologs locate a single mismatch among millions of correctly paired nucleotides has been studied extensively by bulk biochemistry, in vitro atomic force microscopy and single-molecule imaging, and visualizing MutS using in vivo cell biology approaches (7–20). In vitro single-molecule studies largely indicate that MutS operates as a searching clamp diffusing along DNA in a 1D search process (8, 10, 13). In this model, after mismatch recognition, MutS can dwell at the mismatch before exchanging ADP for ATP, converting into a stable ATP-bound sliding clamp, and subsequently, diffusing away from the mismatch at a faster rate in search of MutL and possible strand discrimination signals (8, 10, 15). Studying the mechanistic steps of the search process with DNA curtains has provided evidence that MutS may identify errors through a combination of 1D sliding and a 3D pathway (15). These results show that MutS predominantly searches DNA by 1D diffusion but is also capable of mismatch recognition through a 3D pathway, in which MutS binds the mismatch without engaging in a prior 1D search. The 3D pathway for mismatch recognition has been proposed to help MutS circumvent protein barriers that would exist in vivo. Additional analysis of MutS on DNA curtains suggests that DNA binding proteins largely prevent 1D diffusion by MutS homologs (14). These results indicate that a search mechanism limited to 1D diffusion would present a significant challenge for identifying replication errors in vivo. Therefore, it is still not clear how MutS searches for errors within the supercoiled, compacted, and protein-laden nucleoid in the crowded 3D environment of a living bacterial cell.

Each of these in vitro single-molecule studies has provided important insight into the target search dynamics of MutS, and these studies and other bulk studies have elucidated many of the mechanistic steps leading to mismatch identification on naked DNA in vitro. Of course, most of these studies have been limited to analysis of MutS in isolation on protein-free DNA; they often do not incorporate replication proteins or other in vivo obstacles that are likely to impact the in vivo search process. The search dynamics of MutS have not been investigated at single-molecule resolution for any MutS homolog in live cells.

The need to understand how the biochemistry of MutS homologs translates into mismatch identification in live cells has led to several studies across many organisms using a variety of approaches, including fluorescent protein fusions and bulk fluorescence imaging. These studies have shown that MutS homologs form foci that colocalize with the DNA replication machinery (replisome) and that interaction of MutS with replication processivity clamps [proliferating cell nuclear antigen (PCNA) or β-clamp] is important for focus formation (16, 17, 19–22). In Saccharomyces cerevisiae, MMR is also temporally coupled to DNA replication (23). However, the colocalization between bacterial MutS and the replisome is far from absolute. For example, in the absence of an exogenous mutagen, bulk microscopy detects fluorescent MutS foci in only ∼9% of B. subtilis cells. Furthermore, only about one-half of these 9% of cells with a MutS focus show colocalization of MutS to the replisome (20). Therefore, only about 4.5% of cells under normal growth conditions in B. subtilis show colocalization between MutS and the replisome (20). Obviously, the behavior of MutS in the remaining 91% of cells without a MutS focus has been impossible to determine using ensemble fluorescence techniques.

Most of the MutS pool cannot be visualized by bulk microscopy, because the large fluorescent foci that are visible by ensemble fluorescence imaging account for only 10% of cellular MutS (that is, ∼8 of 80 dimers present in a cell) (24). Therefore, prior studies have been unable to image and quantify the behavior of MutS outside of large static foci visible during ensemble fluorescence microscopy. Therefore, to understand MutS location and dynamics in vivo, only a more sensitive method with higher spatial and temporal resolutions can unambiguously determine the extent to which MutS is enriched near the replisome and answer two open questions. Is MutS enrichment at the replisome constitutive in bacteria? Does MutS search for mismatches away from the replisome in vivo?

To further understand the process of MMR inside cells and gain insight into the movement and location of MutS molecules in vivo throughout MMR, we probed the spatial distribution, dynamics, and genomic distribution of MutS in live bacteria with superresolution fluorescence imaging (25–27), single-molecule tracking (28–30), and ChIP-sequencing (ChIP-seq) approaches. We investigated the effect of replisome association, active DNA synthesis, mismatch binding, nucleotide binding, and presence of MutL on the dynamics of MutS movement in vivo. This study applies single-molecule imaging to a dedicated DNA repair pathway and captures the trafficking behavior of MutS, a protein that is conserved from bacteria to humans. A high degree of spatial and temporal resolution has allowed us to view the dynamic movement and search process of single MutS molecules during MMR in vivo.

Results

Localization and Dynamics of MutS in Live B. subtilis.

We constructed B. subtilis strains natively expressing MutS fused to the photoactivatable fluorescent protein PAmCherry1 as the sole source of MutS (Fig. 1A). MutS-PAmCherry retains MMR activity, and the fusions are stable against proteolytic loss of PAmCherry (SI Appendix, Fig. S1). Because B. subtilis is extremely susceptible to fluorescent impurities present in the environment, producing spurious fluorescent spots that mimic real PAmCherry signals, the imaging sample was carefully prepared for each experiment by cleaning and bleaching to avoid identifying false signals (SI Appendix, SI Text and Fig. S2). To investigate the dynamics of each MutS copy with subdiffraction resolution (<35 nm), a 405-nm laser was used at low power to produce one to three copies of emissive PAmCherry per cell at a time (Fig. 1B and Movie S1). The MutS-PAmCherry copies in this photoactivated subset were imaged and tracked in real time until the PAmCherry molecules were photobleached. Through 10–30 iterations of this photoactivated localization microscopy activation/imaging/photobleaching cycle, superresolution images were constructed, and multiple single-molecule trajectories were recorded for each cell (31–34). In cells expressing MutS-PAmCherry, MutS explores the entire cell while also experiencing significant enrichment and confinement near the cell center or cell quarter positions (Fig. 1C), where the replisome has been shown to reside in previous studies (35, 36).

Fig. 1.

Location and dynamics of MutS in live B. subtilis. (A) Labeling scheme for MutS-PAmCherry. RBS, ribosome binding site. (B) Representative frames showing the photoactivation of a single copy of MutS-PAmCherry in a cell. Purple and green lines above the frames correspond to the photoactivation laser pulse and the imaging laser. (C, Lower Left) Photoactivated localization microscopy (PALM) reconstruction and (C, Right) single-molecule trajectories of MutS-PAmCherry in (C, Upper Left) a live B. subtilis cell. Each subdiffraction-limited coordinate of MutS-PAmCherry is plotted in the PALM image as a Gaussian blur with width equal to its localization uncertainty. The red arrow indicates a region of MutS accumulation. White dashed lines indicate the computer-detected cell boundary. (Scale bars: 1 μm.)

Localization and Dynamics of the DNA Replication Machinery in Live B. subtilis.

To visualize the replisome position as a control and determine whether the sites of MMR and DNA replication coincide, we labeled DnaX, part of the processivity clamp loader and a proxy for the replisome, with the yellow fluorescent protein mCitrine and expressed DnaX-mCitrine ectopically under control of a xylose-inducible promoter. This inducible promoter allowed control of DnaX-mCitrine expression by adjusting the concentration of xylose in the growth medium. Expression of DnaX-mCitrine did not affect the growth rate of the cells (SI Appendix, Fig. S3). Consistent with earlier studies (36), we found that DnaX forms clusters at either the midcell or the one-quarter cell positions (Fig. 2A). Because each replisome contains multiple copies (37) of DnaX-mCitrine with overlapping fluorescence signals (Fig. 2A) and because existing algorithms designed to extract single-molecule positions from images of densely populated fluorophores are not capable of accurately recovering positions if multiple emitters are separated from each other by less than 100 nm (38, 39), we used a photobleaching-assisted approach (40, 41) to achieve single-molecule localization of DnaX (Fig. 2B). On identifying mCitrine molecules that photobleach from one imaging frame to the next, the mean of the frames after photobleaching was subtracted from the mean of the frames before photobleaching to produce the point spread function of the photobleached molecule. The position of this point spread function was subsequently determined from a fit to a 2D Gaussian function. We found that, within a cluster, DnaX molecules are, on average, separated from each other by 54 (perhaps corresponding to replication forks in a replisome early in the replication cycle) or 119 nm (potentially corresponding to replication forks in a replisome during the later stage of the replication cycle) (Fig. 2C) (36). In addition, from time-lapse imaging without photobleaching (Movie S2), we also observed that DnaX clusters are neither mobile nor strictly stationary; instead, they engage in subtle motions exploring a small domain of size 84 ± 20 (SD) nm as measured by the radius of gyration of the centroid position (Fig. 2D). The subcellular localization and dynamics of DnaX resemble those of the β-clamp protein (DnaN) (SI Appendix, SI Text and Fig. S4). Because the DnaN-mCitrine fusion partially compromises MMR activity in vivo, throughout this study, we use DnaX-mCitrine as a marker for the replisome position. Overall, our results support models describing the B. subtilis replisome as a confined assembly (36, 42) as opposed to the model of a mobile replisome complex that tracks on chromosomal domains as described for Escherichia coli (43).

Fig. 2.

Positioning of the replisome. (A) Sample fluorescence image of DnaX-mCitrine in 5 cells and localization probability of DnaX in 161 cells along the longitudinal cell axis (Inset). The replisome appears most frequently at the one-quarter positions in exponential-phase cells. (B) Photobleaching-assisted localization of single DnaX-mCitrine molecules within a cluster. The intensity of a cluster is plotted against time (20 ms per frame), and photobleaching events are identified by maximum likelihood estimation. The point spread function (PSF) images of photobleached single molecules can then be obtained by subtracting the average intensity of the frames after the photobleaching from that of preceding frames (schematic representation shown in Inset), allowing the precise location of the photobleached molecule to be determined through 2D Gaussian fitting. Representative PSFs for two photobleached mCitrine molecules are shown below the cluster intensity trajectory. (C) Distribution of separation distance between DnaX-mCitrine within a replisome as determined from photobleaching-assisted localization, with two sample overlapping PSFs shown. (D, Left) The radii explored by DnaX-mCitrine as calculated by tracking the motion of each cluster centroid using low-power time-lapse imaging and (D, Right) the size distribution of the domain explored by each cluster, illustrating the subtle replisome motion that explores small domains of ∼84 nm in radius on average. (Scale bar: 1 μm.)

Relative Positions and Dynamics of DnaX-mCitrine and MutS-PAmCherry.

Imaging cells expressing both DnaX-mCitrine and MutS-PAmCherry (Movie S3), we found that MutS accumulates near the replisome, despite being overall more mobile than DnaX (Fig. 3A). The instantaneous speed from single-molecule MutS tracks (44) also shows dependence on the separation distance between MutS and the replisome. On entering the replisome region (separation distance <100 nm), MutS slows down to match the average speed of DnaX clusters (Fig. 3B), likely as a result of the known direct interaction between MutS and the replication processivity clamp (β-clamp) (18, 20) or because of MutS engaging in mismatch searching and binding on replisome-proximal DNA (18, 20) or a combination thereof. Cross-correlation analysis shows that MutS instantaneous speed correlates positively with the separation distance from the replisome (Fig. 3C). However, the relatively low correlation amplitude of ∼0.2 indicates heterogeneity among MutS subpopulations (that is, some MutS do not slow down or slow down only transiently when passing by the replisome). To quantify how much time a MutS protein spends within the replisome region, we fit the cumulative probability of the dwell time of MutS in the replisome region [P(t) > t] with a two-term exponential decay function (Fig. 3D) and obtained two dwell time constants of 25 (42%) and 188 ms (58%). These dwell time constants represent a lower bound, because we only analyzed single MutS trajectories that start outside the replisome, remain trackable within the replisome, and end outside the replisome (Fig. 3D, Inset). As a result, photobleaching and blinking of the fluorophore do not affect the dwell time analysis, but only MutS trajectories that start and end outside the replisome were taken into account. We conclude that the majority of MutS molecules that pass by the replisome (58%) dwell there for a considerable time (at least 188 ms), likely to engage in a local search for mismatches, whereas the other population (42%) is freely diffusing within the cell (SI Appendix, Fig. S5).

Fig. 3.

Two-color imaging results from single cells expressing both MutS-PAmCherry and DnaX-mCitrine. (A) Photoactivated localization microscopy reconstruction of MutS-PAmCherry (magenta) in a cell (Top) with MutS-enriched regions indicated with white arrowheads and (Middle) overlaid with DnaX-mCitrine clusters (green). A representative time-coded trajectory showing MutS entering, dwelling at, and leaving one of the replisome regions is shown in Bottom. (Scale bars: 1 μm.) (B, Upper) Separation distance from the replisome and (B, Lower) instantaneous speed as a function of time for the MutS trajectory shown in A. Gray indicates the time spent in the replisome region. The red curve indicates a prolonged period of decreased MutS speed. Black dashed lines indicate (Upper) 100-nm MutS-DnaX separation distance and (Lower) average DnaX speed (0.5 μm/s as measured by tracking cluster centroids). (C) Cross-correlation between the separation between MutS and the center of DnaX cluster and the instantaneous speed of MutS from 29 trajectories in 11 cells normalized from −1 to 1 (66). Error bars represent SEM. (D) Cumulative probability distribution of the time period that MutS (red “S”) spends within the same replisome region (blue “R”) fit to a two-term exponential decay function (dashed line) P = A₁exp(−t/τ₁) + A₂exp(−t/τ₂), where A₁ = 0.42, A₂ = 0.58, τ₁ = 25 ms, and τ₂ = 188 ms. The error bars are SDs from seven rounds of bootstrapping. The dwell time distribution is constructed using 751 trajectories at least 10 frames long for molecules that can be tracked from the time that they enter the replisome region until they leave the replisome (Inset).

MutS Accumulates at the Replisome, Regardless of Mismatch Formation Rate.

It is necessary to sample a large number of cells to make meaningful conclusions, because there are only ∼160 copies of MutS-PAmCherry in each cell, and PAmCherry photoactivation efficiency is 4–50% (45, 46). Furthermore, only some small fraction of the imaged cells has mismatches to which MutS is able to respond (Fig. 4E). Therefore, to compare intracellular DnaX-mCitrine and MutS-PAmCherry positions across many cells, we plotted the percentage of DnaX and MutS localizations at corresponding positions inside a normalized cell as probability density maps (SI Appendix, Fig. S6) (47). Only cells with two replisome clusters were used to construct the density maps, and chain-forming cells were discarded from additional analysis. A density map of WT cells treated with the mismatch-forming drug 2-aminopurine (2-AP; WT+) (Fig. 4E and SI Appendix, SI Text) shows that regions with the highest DnaX densities are also those most populated by MutS (Fig. 4C). Importantly, the same pattern was also found in WT cells without 2-AP (WT−) (Fig. 4 A and compare B with C). Although previous bulk fluorescence studies of B. subtilis have shown that <10% of WT− cells form MutS foci near the replisome (16, 20), here at the single-molecule level with improved sensitivity to capture transient dwelling behaviors (Fig. 3B), we reveal that the enrichment of MutS near the replisome is much more prevalent than previously concluded, even in cells with only the very low natural mismatch formation rate (i.e., no 2-AP addition). Furthermore, this mismatch-independent recruitment of MutS in B. subtilis resembles the behavior of MutSα in S. cerevisiae cells observed using bulk fluorescence (19), suggesting that mismatch recognition is an integral component of replisome function and has been conserved from bacteria to eukaryotes. However, we show that bacterial MutS is highly dynamic, displaying heterogeneous motion in addition to its replisomal association.

Fig. 4.

MutS localization and dynamics in WT cells. (A) Photoactivated localization microscopy (PALM) reconstruction (magenta) and single-molecule trajectories (red) of MutS-PAmCherry overlaid with DnaX-mCitrine (green and blue) and phase-contrast cell images. Overlapping signals are colored in white. Orange arrows indicate replisome regions at which preferential MutS enrichment or dwelling is observed. (Scale bar: 1 μm.) (B and C) Localization probability density maps of (Upper) DnaX-mCitrine (blue-green) and (Lower) MutS-PAmCherry (red-yellow) within a normalized cell. White lines designate the one-quarter, one-half, and three-quarters positions along the cell long axis and the one-half position in the transverse direction. In total, (B) 108 WT− and (C) 91 WT+ cells with two replisome clusters were used to generate the corresponding density maps. To allow for quantitative comparison of colocalization between different cases, the Pearson correlation coefficients between each pair of DnaX/MutS density maps were calculated. The correlation coefficients for WT− and WT+ cells are 0.83 and 0.81, respectively. Grid pixel size: ∼100–200 nm. (D) Diffusion coefficients of MutS-PAmCherry as a function of separation distance from the nearest replisome. Error bars indicate 95% confidence interval. The three arrows point to the 50-, 200-, and 400-nm separation distances (from left to right) at which the MutS diffusion coefficient distributions are further analyzed in F. (E) Distribution of the probability of a mismatch occurring in an observed cell under normal growth condition (blue) and 2-AP treatment (red) over time. The vertical purple dashed line indicates the average duration (210 s) of observation for each cell in PALM experiments. (F) Distribution of the MutS diffusion coefficients at three distances away from the replisome. Over 3,000 trajectories were analyzed for both WT− and WT+ cells.

MutS Speed Increases After Mutagen Treatment.

To understand the spatial dependence of MutS motion, we analyzed the average diffusion coefficient of MutS as a function of separation distance from the nearest replisome (Fig. 4D). The diffusion coefficient, D, was calculated from the mean square displacement, and only data from the first one-quarter of the time lags for each of over 3,000 trajectories longer than 10 frames were analyzed to minimize errors associated with higher time lag values (48). Under both normal and mutagenic growth conditions, the MutS diffusion coefficients are on the same order of magnitude as those of DNA ligase and PolI in live E. coli (49) and those of DNA-bound proteins exhibiting a combination of 1D and 3D motions in other in vivo systems (50, 51). However, despite the nearly identical net MutS localization patterns for WT− and WT+ cells (Fig. 4 B and C), on the single-protein level, we found that MutS exhibits an overall faster motion in WT+ cells compared with WT− cells (Fig. 4D). This in vivo difference in speed is consistent with in vitro observations that MutS switches from rotation-coupled sliding to a faster rotation-free sliding after mismatch binding (8, 10, 15).

Because the diffusion coefficient, D, in Fig. 4D only reflects an arithmetic mean of all MutS molecules, we probed the distributions of D for MutS at various distances away from the replisome. In both WT− and WT+ cells (Fig. 4F), the diffusion coefficient distribution for MutS near the replisome (50-nm separation) (red curves in Fig. 4F) is bimodal, with a slow population (D < 0.1 µm²/s) and a faster population (D ∼ 0.15–0.2 µm²/s). The WT+ distribution is different in two notable ways: 2-AP causes the slow population to become even slower, presumably because of significant numbers of MutS that have just recognized mismatches and are, thus, slower, and 2-AP significantly increases the proportion of MutS within the faster population for both replisome-proximal MutS and MutS farther away (i.e., 200- and 400-nm separation), corresponding to postmismatch binding MutS undergoing the faster, rotation-free sliding.

Regardless of whether 2-AP is present, we observe that MutS slows as it approaches the replisome in both WT+ and WT− cells, suggesting that error searching and possibly, subsequent binding events are restricted to nascent DNA in the neighborhood of DNA replication sites and likely initiated by interactions between MutS and the replisome.

MutS/Replisome Interaction Is Necessary for MutS Recruitment and MMR in Vivo.

To further understand the relationship between DNA replication and the position and dynamics of single MutS molecules, we constructed four strains, each designed to impair one of four MMR steps: (i) MutS binding to β-clamp, (ii) mismatch recognition, (iii) MutS nucleotide binding, and (iv) subsequent MutL recruitment (Fig. 5A). We tested the effect of replisome interaction perturbation on MutS motion and location using a two-pronged approach. First, we examined MutS800, a MutS variant with the domain that has β-clamp affinity removed. MutS800 is able to bind mismatches (18) and maintains similar ATPase activity to full-length MutS (SI Appendix, Fig. S7). Second, we complemented the MutS800 investigations with studies of the β-clamp variant dnaN5, which is compromised for interaction with MutS (52). Relative to the WT− and WT+ density maps, the MutS localization pattern is drastically changed in untreated MutS800 (MutS800−) cells (compare Fig. 5C with Fig. 4 B and C). Although DnaX locations remained largely the same, MutS800 was uniformly distributed throughout the cell, lowering the Pearson correlation between DnaX and MutS800 from 0.83 in WT− cells to 0.50 in MutS800−. The significant decrease in preferential enrichment of MutS near the replisome confirms that the recruitment of MutS to the replisome observed in WT cells depends, in part, on interactions between MutS and the β-clamp.

Fig. 5.

Response of MutS to sequential blocking of MMR steps. (A) Schematic diagrams showing the first four steps of MMR, including replisome binding, mismatch recognition, ATPase activity, and MutL recruitment, each of which is blocked in one of four mutant strains. (B, F, and J) Two-color images of representative cells from MutS800, MutS[F30A], and MutS[K608M] strains. (C, G, K, and N) Localization probability density maps of untreated cells for each strain generated from N cells with two replisome clusters. Pearson correlation coefficients between DnaX and MutS densities are listed in the lower right corner of corresponding MutS density maps. Note that MutS density maps may exhibit similar intensity levels but have different correlation coefficients because of differences in the positioning of corresponding DnaX density maps. (D, H, L, and O) Density maps of 2-AP–treated cells. (P) Density maps for 2-AP/HPUra double-treated cells from the ΔmutL+ strain. HPUra restores MutS enrichment around the replisome and also causes simultaneous shifting of DnaX and MutS localizations toward cell center. (E, I, M, and Q) Diffusion coefficients of MutS-PAmCherry variants as a function of separation distance from the nearest replisome. Error bars indicate 95% confidence interval.

The density maps after 2-AP treatment (Fig. 5D) suggest that MutS800 still partially responds to mismatch incorporation (the correlation coefficient increases from 0.50 to 0.70 with addition of 2-AP). Although MutS800 is still capable of binding nucleotide (SI Appendix, Fig. S8) and mismatched DNA in vitro (18), only 10–12% of MMR activity is retained in MutS800 cells (SI Appendix, Fig. S1). This lack of MutS800 density further shows the importance of replisomal interactions for efficient mismatch recognition in vivo. We interpret the slightly higher densities near the one-quarter positions of the MutS800+ density map (Fig. 5D) as MutS800 engaged in the residual 10–12% of MMR activity. In addition, the diffusion rates for MutS800 were significantly increased relative to WT− and WT+ (Fig. 5E). One possible contributing factor to the increase in diffusion rate is that mismatch-stimulated ATPase activity of MutS800 is twofold diminished relative to that of the WT MutS, although the basal ATPase activity between the two proteins in the presence of homoduplex DNA was the same (SI Appendix, Fig. S7). However, our results for the complementary experiment in the DnaN5 variant strain also show a partially compromised colocalization pattern between MutS and DnaX and a faster diffusion profile for MutS in the DnaN5 strain compared with WT cells (SI Appendix, Fig. S9). These results indicate that the increase in MutS800 diffusion should not be solely attributed to a slight decrease in stimulation of ATPase activity in response to a mismatch. One potential explanation for this change in dynamics is that the MutS–β-clamp interaction facilitates MutS binding to DNA, and in the absence of such an interaction, MutS cannot efficiently engage in slow 1D searching motion. Because of the significantly increased diffusion rate of MutS800, it is likely that the 10–12% of residual MMR activity in this variant is achieved by mismatch recognition on 3D diffusion.

MutS Recruitment to the Replisome Occurs Independently of Mismatch Recognition.

Because MutS positioning did not change on 2-AP treatment in WT cells, we tested whether recruitment of MutS to the replisome is contingent on mismatch binding. We analyzed the distribution and motion of MutS[F30A], which is unable to recognize mismatches (24). Both with and without 2-AP, this mutant preserved the elevated MutS density around the replisome observed in WT cells (compare Fig. 5 G and H with Fig. 4 B and C). Also, the dependence of diffusion rates on separation distance (Fig. 5I) for both MutS[F30A]− and MutS[F30A]+ was virtually identical to that of WT−. Thus, as expected, because MutS[F30A] cannot bind mismatches (20), all aspects of MutS[F30A] localization and motion remain unresponsive to mismatches caused by 2-AP (Fig. 5 G–I). The highly similar positioning and dynamics of MutS[F30A]− and MutS in WT− cells support the notion that the recruitment of MutS by the replisome not only precedes mismatch recognition but also, occurs independently of it. This dynamic is consistent with the ability of MutS to efficiently respond to very rare mismatches (2), because it is enriched at the replisome, in close proximity to potential mismatches and before errors occur. The recruitment of MutS to the replisome also ensures that MutS has access to newly replicated, “naked” DNA strands along which prolonged sliding by MutS would be possible.

Nucleotide Binding Is Necessary for MutS Recruitment.

Another MMR step that is poorly understood in bacterial cells is how nucleotide binding affects the search phase, movement, and localization of MutS homologs in living cells (53–55). Abundant in vitro evidence shows that MutS carries out an ATPase cycle, in which it is able to engage in mismatch searching while bound to ADP, and on mismatch detection, MutS binds ATP, inducing a conformational change to form a stable clamp capable of more rapid 1D diffusion along the DNA (8, 11, 55). To better understand how the ATPase cycle affects MutS localization and dynamics in vivo, we constructed a strain expressing MutS[K608M]-PAmCherry as the only source of MutS. MutS[K608M] in B. subtilis is the homologous substitution to E. coli MutS[K620M], which has a far reduced affinity for nucleotide and therefore, little ATPase activity (56). The ability of E. coli MutS[K620M] to bind mismatched DNA is not clear. One study shows mismatch binding by this MutS variant (56), whereas another does not (57). In B. subtilis, MutS[K608M] changes the highly conserved Walker A motif in MutS, and purified MutS[K608M] has no measurable ATPase activity or nucleotide binding (SI Appendix, Figs. S7 and S8) (56). Strikingly, MutS[K608M] displayed highly diffusive behavior both with (MutS[K608M]+) and without (MutS[K608M]−) 2-AP treatment (Fig. 5 J–L). MutS[K608M] motion was also unresponsive to 2-AP treatment, and similar to MutS800, the diffusion of MutS[K608M] did not depend on its separation distance from the replisome (Fig. 5M). However, in contrast to MutS800 motion, which was significantly faster, this constant MutS[K608M] diffusion rate throughout the cell closely resembled the rate of WT+ MutS at positions >250 nm from the replisome (Fig. 5M). These results suggest that processivity clamp interaction alone is not entirely sufficient to recruit MutS to the site of DNA synthesis, and proper nucleotide binding by MutS is also necessary for positioning of MutS within the cell, possibly to return MutS to its mismatch searching state near the replisome. Importantly, regardless of whether or not MutS[K608M] is able to recognize mismatches as efficiently as WT MutS, MutS[F30A], which is completely deficient in mismatch recognition, localizes to the replisome. These results for MutS[K608M] and MutS[F30A] lead us to the conclusion that replisome interaction and nucleotide binding, but not mismatch detection, are necessary for MutS positioning to the replisome.

MutS Only Recognizes Mismatches Spatially Close to the Replisome.

After mismatch detection, MutS binds MutL to form a complex (24), which is proposed to then slide away from the mismatch in search of MutL and strand discrimination signals along the DNA (15, 16, 18). To further test whether MutS binds mismatches near or distal to the replisome, we probed the effect of MutL binding on MutS dynamics by imaging DnaX and MutS in a ΔmutL strain having no MMR activity. Without 2-AP (ΔmutL−), the localization of MutS in ΔmutL largely resembles that observed in the WT and MutS[F30A] strains (compare Fig. 5N with Fig. 4 B and C and Fig. 5 G and H), indicating that the preloading of MutS at the replisome is unaffected by the absence of MutL. However, it is notable that, when mismatches were induced in ΔmutL cells by 2-AP treatment (ΔmutL+), the density of MutS in replisome-proximal regions diminished (compare Fig. 5O with Fig. 4C and Fig. 5H). We postulate that the decline in MutS enrichment at the replisome is caused by MutS remaining mismatch-bound in the absence of MutL and thus, being carried away from the replisome with newly synthesized mismatch-containing DNA as DNA replication proceeds. We tested this hypothesis by imaging MutS in the ΔmutL strain incubated with 2-AP for 1 h followed by treatment with 6(p-Hydroxyphenylazo) uracil (HPUra), which blocks DNA replication (58). Consistent with our hypothesis, enrichment of MutS at the replisome was restored after DNA replication was arrested (compare Fig. 5 O with P), showing that, in ΔmutL, MutS is a marker of mismatch positions on the newly replicated DNA and that, in the absence of HPUra, the attenuated MutS accumulation observed in ΔmutL+ (Fig. 5O) is caused by the mismatch-bound MutS being carried away from the replisome during ongoing DNA synthesis. The process of MMR from mismatch detection by MutS to ultimate replacement of the error-containing strand of DNA must occur quickly, because deletion of MutL was necessary to observe an effect of 2-AP treatment on MutS position (Fig. 5 N and O).

Overall, the dependence of mismatch recognition on replisome coupling, as observed in MutS800 and DnaN5 strains, is further supported by the MutS localization in ΔmutL. Here, DNA mismatches will have accumulated throughout the genomic DNA (Fig. 5O) before HPUra treatment because of prolonged 2-AP exposure without MutL. Therefore, if MutS could bind mismatches away from the replisome, the MutS distribution would resemble the diffuse pattern observed in ΔmutL+ cells (Fig. 5O). Rather, HPUra causes MutS to resume its normal enrichment at the replisome (Fig. 5P). This restoration of MutS density shows that the proximity of mismatched DNA to the replisome plays a fundamental role in MMR initiation, because mismatches distal to the replisome are not efficiently targeted by MutS. Together with the loss of MMR activity in MutS800 and the partial loss of MMR in DnaN5 (18) (SI Appendix, Fig. S1), this observation further indicates that proximity between MutS and the replisome is critical for MutS to efficiently locate mismatches in vivo. In contrast to WT cells, MutS slows down on 2-AP treatment in ΔmutL (Fig. 5Q), which presumably results from the increased number of mismatch-bound MutS molecules distal to the replisome.

In contrast to treatment with 2-AP, which had no effect on replisome localization, blocking DNA replication with HPUra changed the locations of both the replisome and MutS. For predivisional cells with two DNA replication sites, the replisome and MutS were both shifted inward from one-quarter positions toward the cell center (Fig. 5P). This shifting of the replisome and MutS inward is likely due to the fact that, with DNA replication paused, cells continue to grow but fail to synthesize DNA. The simultaneously shifted colocalization pattern observed in HPUra-treated cells again highlights the functional correlation between DNA replication and repair.

MutS Interacts with Newly Replicated DNA and Essential DNA Polymerases Both in Vivo and in Vitro.

The experiments described above all use fusions of fluorescent proteins to MutS or DnaX. To biochemically test whether untagged MutS physically associates with replicating DNA in vivo, we synchronized cells for replication initiation using a temperature-sensitive variant of the replicative helicase loader DnaB (DnaB134) followed by ChIP-seq of MutS and the essential catalytic DNA polymerase subunits PolC and DnaE to determine the location of each protein on the chromosome (59, 60). The B. subtilis genome consists of a single circular chromosome with a single origin of replication (oriC). DNA replication commences at oriC and proceeds bidirectionally toward terC. In synchronized culture during preinitiation, we observed little to no enrichment of MutS, DnaE, or PolC on the chromosome (SI Appendix, Fig. S10). In contrast, 10 min after replication has commenced, PolC, DnaE, and MutS are coenriched at the site of replication initiation (oriC), each displaying about twofold enrichment (Fig. 6A); 25 min after initiation of replication, PolC, DnaE, and MutS have migrated farther from oriC, and coenrichment of all three proteins is observed (SI Appendix, Fig. S10). MutS enrichment profiles correlate well with those of DnaE and PolC, because the Pearson correlation coefficients are 0.70 and 0.85 for MutS/DnaE and MutS/PolC, respectively, at 10 min (Fig. 6A). After 25 min, the correlation coefficients are 0.82 and 0.92 for MutS/DnaE and MutS/PolC, respectively (SI Appendix, Fig. S10). The enrichment is quite broad, occupying a region over 200-kbp wide at 10 min (Fig. 6B) and a region over 1,000-kbp wide at 25 min (SI Appendix, Fig. S10). The amounts of enrichment near oriC at 10 min and more widely at 25 min are, therefore, substantial given the broad distribution of each protein on the DNA. Importantly, there was little enrichment in ChIP-seq of MutS800 at 10 min after replication initiation, consistent with the MutS800 variant being deficient in recruitment to the replisome, and ChIP-seq using antibodies directed against MutS in a strain lacking mutS also yielded very little enrichment (Fig. 6A). Given these controls, the less than 80% synchrony within our system (61), and the very similar patterns of enrichment that we observed in our independent ChIP-seq of two essential components of the B. subtilis DNA polymerase, we are confident that the broad enrichments observed at 10 and 25 min after initiation of replication represent the bona fide location of the DNA polymerase and MutS on the chromosome in the plurality of the cells in our synchronized culture. We conclude that MutS is physically associated with the site of ongoing DNA synthesis and rapidly loaded at oriC on chromosomal replication initiation. MutS is then able to dynamically monitor newly replicated DNA for mismatches at or near the replisome throughout the DNA replication cycle.

Fig. 6.

MutS recruitment to DNA on replication initiation and interaction with DNA polymerase subunits. (A) Analysis of pooled ChIP-seq data from two independent experiments showing the enrichment levels of MutS and the polymerases DnaE and PolC along the chromosome 10 min after DNA replication initiation. Pearson correlation coefficients (r values) are shown for genome-wide ChIP-seq enrichment profiles of DnaE/MutS and PolC/MutS. The position of oriC is indicated by an arrow above the plots. (B) Overlaid enrichments of MutS, PolC, and DnaE from A in a window of the genome near oriC. (C) Co-IP of DnaE and PolC with MutS using affinity-purified antiserum directed against MutS. Lane 1, 5% input; lane 2, anti-MutS immunoprecipitation.

Because MutS position in space and on the genome correlates very well with that of the replisome, we determined whether MutS interacts with subunits of the replisome other than β-clamp. Using a far-Western blot, we detected direct in vitro interaction between MutS, DnaX, and δ (components of the clamp loader complex) and between PolC and DnaE (catalytic subunits of the B. subtilis DNA polymerase) (SI Appendix, Fig. S11). To test whether untagged MutS physically associates with the replisome in vivo, we carried out coimmunoprecipitation (co-IP) and found that MutS binds both PolC and DnaE in vivo in untreated cells (Fig. 6C). On 2-AP treatment, PolC and DnaE no longer coimmunoprecipitate with MutS (Fig. 6C). We conclude that MutS interacts with the replisome before mismatch formation. Because we carried out the co-IP with the reversible membrane-permeable cross-linker dithiobis(succinimidyl propionate) (DSP), it is possible that the co-IP of PolC and DnaE with MutS occurs indirectly by MutS interaction with β-clamp, but taken together with our imaging and ChIP-seq results, these data suggest that MutS is capable of searching DNA in extremely close proximity to the actively replicating DNA polymerase complex, even in the absence of 2-AP. This coenrichment occurs whether the interaction of MutS with PolC and DnaE in vivo is direct or indirect, and with 2-AP in the growth medium, MutS scans newly replicated DNA in clamp zones trailing the replisome.

Discussion

In the crowded cellular environment, 3D diffusion alone is too slow to allow for efficient detection of base-pairing mistakes. Previous in vitro studies have shown that 1D sliding is another mechanism by which MutS can locate DNA mismatches (15). To further our understanding of the corresponding process from the in vivo perspective and complement and clarify our existing knowledge of MMR across species, we performed single-molecule superresolution microscopy in live B. subtilis. With nanometer-scale spatial resolution and millisecond-scale temporal resolution, we directly visualized and quantified the behavior of MutS in vivo in real time. We find that MutS molecules move to and from the replisome even under normal growth conditions, with exceedingly low mismatch formation rates. When MutS reaches the replisome, it dwells for at least ∼200 ms before releasing and exploring the rest of the cell. MutS is also highly dynamic throughout the cell, a behavior that has not been observed before. The measured diffusion rate for MutS close to the replisome is consistent with that of typical DNA binding proteins undergoing 1D motion in vivo, supporting the hypothesis that 1D sliding is the primary mechanism used by MutS for efficient mismatch recognition near the replisome (62). We also find that the diffusion rate of MutS increases with separation distance from the replisome and after mismatch binding. We show that mismatch detection by MutS must occur near the site of DNA replication and that MutS remains associated with mismatches until MutL is recruited. Furthermore, we show that MutS is loaded at the origin with the replisome in synchronized cells and tracks with the replisome throughout the DNA replication cycle. Our investigation of protein–protein interactions uncovers a physical interaction between MutS and several replisome components, including the two essential DNA polymerase catalytic subunits PolC and DnaE, providing additional evidence that MutS interaction with the replisome is not simply restricted to processivity clamps. We also show by complementary experimental approaches that MutS is enriched along the genome with the DNA polymerase. Moreover, our study provides insight into MutS search dynamics in live cells by using a single-molecule view of MutS during the search for mismatches. We reveal a highly dynamic process, where MutS molecules constitutively move to and from the replisome rapidly and dwell at the replisome for short intervals while searching for errors.

Some of the MutS diffusion coefficients measured in vivo in this study differ from those observed by in vitro FRET experiments (10). We observed MutS diffusion faster than 0.2 μm²/s at sites distant from the replisome region (≥200 nm) for WT cells treated with 2-AP (WT+). Similar in vivo diffusion coefficients and similar discrepancies in diffusion coefficients measured in vivo and in vitro have also been reported, and it has been suggested that the higher diffusion coefficients measured in vivo likely result from the fact that in vivo experiments simultaneously track proteins engaged in 1D and 3D motions. Thus, we propose that detection of a mismatch in vivo is followed by the formation of a stable, quickly sliding ATP-bound MutS clamp as has been observed in vitro (8, 15).

Our results reveal that the heterogeneous behavior of MutS depends on mismatch binding state as well as intracellular location, MutL expression, and nucleotide binding by MutS. We show that nucleotide binding is necessary for the MutS cycle within the cell and that, if MutS cannot engage in nucleotide binding, it is unable to properly localize to the replisome region in vivo. Because MutS[F30A] is recruited to the replisome, we can effectively rule out the possibility that any potential defect in mismatch recognition by MutS[K608M] is responsible for its change in diffusion coefficient or localization. We show that there is a highly dynamic and transient interplay between MutS and the replisome that positions MutS to sites of ongoing DNA replication before mistakes occur, such that MutS can constantly monitor the newly synthesized DNA. Such behavior is somewhat similar to observations with MutSα in eukaryotic cells (19), suggesting that replisome association is a highly conserved mechanism for mismatch detection across species. However, unlike MutSα, which may be able to bind mismatches independent of the replisome (19), our results obtained from B. subtilis suggest that mismatch binding by MutS must occur at the replisome. One possible explanation for this difference in mechanism between bacteria and S. cerevisiae is that bacterial DNA replication occurs continuously, each replication fork moves at a rate of ∼500 nt/s, and in rich growth conditions, multiple rounds of replication initiation can occur before a single-cell division. Conversely, in eukaryotic cells, DNA replication occurs in S phase, the replication fork moves at ∼27 nt/s, and the behavior of biochemical pathways can be regulated in a cell cycle-dependent manner (63–65). Therefore, mismatches must be detected very quickly in a rapidly proliferating bacterium like B. subtilis, or they will become mutations within mere minutes when the next round of DNA replication duplicates the mismatched DNA. In bacteria, such as B. subtilis, near the replisome is arguably the best place for MutS to scan DNA for mismatches barrier-free, because MutS has access to newly synthesized DNA largely free of proteins and is also able to recognize rare replication errors as they are produced because of the spatial proximity between the two. Therefore, the replisome provides a scaffold that allows MutS to target a single mistake among tens of millions of correctly paired nucleotides in a timely manner, guarding the bacterial genome against mutations that could otherwise have deleterious effects on bacterial growth and fitness.

Materials and Methods

Additional detailed descriptions of the materials and methods used in this study can be found in SI Appendix, SI Text.

Sample Preparation for Single-Molecule Imaging.

B. subtilis PY79 cells were grown at 30 °C in Spizizen's S7₅₀ minimal medium supplemented with 1% (wt/vol) arabinose with a starting OD₆₀₀ ∼ 0.1. We used arabinose as opposed to glucose as the carbon source, because the latter suppresses the expression of DnaX-mCitrine. To induce ectopic expression of DnaX-mCitrine, 0.125% xylose was added to the medium. To cause mismatch formation, 2-AP was added at OD ∼ 0.35 to a final concentration of 0.6 mg/mL, and cells were harvested during exponential phase when OD reached ∼0.55–0.65. When used, HPUra (final concentration of 162 μM) was added to the culture immediately before imaging; 2 µL cell culture was pipetted onto a 1% (wt/vol) agarose in S7₅₀ pad, which was sandwiched between two coverslips that had been cleaned by oxygen plasma (Plasma Etch PE50) for 20 min. The sample was then mounted on the microscope objective for imaging.

Microscopy and Imaging Parameters.

DnaX-mCitrine was first imaged under 7-W/cm², 488-nm laser illumination (Coherent Sapphire 488–50), and then, MutS-PAmCherry1 was photoactivated using a 200-ms 405-nm pulse (Coherent 405–100) and imaged with a 561-nm (Coherent Sapphire 561–50) laser with power densities of 35–110 and 120 W/cm². For photobleaching-assisted DnaX-mCitrine localization, a higher-power density of 70 W/cm² was used. Wide-field single-molecule epifluorescence microscopy was performed on an Olympus IX71 Inverted Microscope. Fluorescence emission was collected by a 1.40-N.A. 100× oil-immersion phase-contrast objective and detected on a 512 × 512-pixel Photometrics Evolve EMCCD at a rate of 25 Hz. The photobleaching-assisted DnaX-mCitrine localization experiment was performed using a frame rate of 50 Hz. Appropriate dual-color dichroic and band-pass filters (Semrock) in the emission pathway rejected scattered laser light and maximized the signal to noise ratio.

ChIP-Seq and ATPase Assay.

Procedures and results are provided in SI Appendix, SI Text.