SMC protein RecN drives RecA filament translocation for in vivo homology search

Significance

Double-strand breaks (DSBs) are a particularly lethal form of DNA damage. In all domains of life, DSBs can be faithfully repaired via homologous recombination. A key step in this pathway involves homology search, where a highly conserved recombinase (RecA/Rad51) associates with the break site to search for its spatially distant homologous partner. How this complex yet efficient search mechanism is carried out inside a cell has been an abiding enigma. In this study, using quantitative live-cell imaging, we uncover a dynamic search process that involves directional translocation of the RecA filament, via action of structural maintenance of chromosome (SMC)-like protein RecN. Taken together, our data point to a robust strategy to enable long-distance homology search.

Abstract

While the molecular repertoire of the homologous recombination pathways is well studied, the search mechanism that enables recombination between distant homologous regions is poorly understood. Earlier work suggests that the recombinase RecA, an essential component for homology search, forms an elongated filament, nucleating at the break site. How this RecA structure carries out long-distance search remains unclear. Here, we follow the dynamics of RecA after induction of a single double-strand break on the Caulobacter chromosome. We find that the RecA-nucleoprotein filament, once formed, rapidly translocates in a directional manner in the cell, undergoing several pole-to-pole traversals, until homology search is complete. Concomitant with translocation, we observe dynamic variation in the length of the filament. Importantly in vivo, the RecA filament alone is incapable of such long-distance movement; both translocation and associated length variations are contingent on action of structural maintenance of chromosome (SMC)-like protein RecN, via its ATPase cycle. In summary, we have uncovered the three key elements of homology search driven by RecN: mobility of a finite segment of RecA, changes in filament length, and ability to conduct multiple pole-to-pole traversals, which together point to an optimal search strategy.

DNA double-strand breaks (DSBs), if left unrepaired (or incorrectly repaired), can lead to loss of genetic information, chromosomal rearrangements, mutagenesis, and even cell death. Homologous recombination, a process that usually ensures error-free repair of DSBs, has emerged as an evolutionarily conserved pathway from bacteria to eukaryotic cells (1). Recent work has shown that recombination-based repair is not only local, such as during postreplicative cohesion, but that even distant homologous regions can participate in such repair (1–11). This is particularly relevant for bacteria, where an extensive postreplicative cohesion period is absent (12–18). Although DSBs occur often at stalled or broken replication forks, breaks can additionally occur independently of replication. Examples include X-rays, IR radiation, chemicals and metabolites such as bleomycin, topoisomerase dysfunction (from inhibitor action for example), processing of closely spaced cyclobutane pyrimidine dimers, transposable element-mediated events, and bacterial toxins that directly cause DSBs (6, 7, 19–23). Several of these breaks rely on recombination-based repair (6, 7, 20, 24, 25). Given the potential lethality associated with unrepaired breaks (6, 8, 19, 20, 25–28), it is vital that there be regulatory mechanisms to sense and initiate, translocate and search, and finally identify even distant homology regions for repair. These strategies for long-distance homology search and repair must potentially depend on the spatial organization of chromosomes—its extent, geometry and dimensionality (29)—and must involve energy transduction.

In bacteria, an extensive homology search program is effected by the highly conserved recombinase, RecA (30, 31). In the first step of the recombination pathway, DSB-ends are processed to reveal single-strand DNA (ssDNA) overhangs, on which RecA can assemble into a filamentous structure (3, 32, 33), which in turn initiates homology search. In vitro studies have shown that RecA molecules in the nucleoprotein filament possess ATPase activity that can influence: 1) formation of a continuous filament during RecA loading on ssDNA (34), 2) growth and shrinkage of the filament (35–37); 3) release of the filament from heterologous pairing (31, 38); and 4) RecA turnover during strand invasion (37–39). In vivo studies on DSB repair have made use of engineered endonuclease-based DSB-inducing systems (such as the I-SceI system) to make targeted breaks on the chromosome and track the associated dynamics (3–5, 10, 11, 40). Using such a system in Escherichia coli to study dynamics of long-distance homology search, imaging of RecA has revealed the presence of large assemblies of RecA, described as RecA filaments (or “bundles”), that can even span the length of the cell (4, 5, 41). How this RecA structure carries out homology search remains unclear.

In addition, several accessory proteins are reported to modulate the association and turnover of RecA with ssDNA (42, 43). Of these, RecN, a highly conserved repair-associated protein, has been shown to interact with RecA and is essential for DSB repair (43–47). RecN belongs to the Structural Maintenance of Chromosome (SMC) family of proteins that play central roles in chromosome dynamics (48). While their function in chromosome organization and segregation in bacteria has been well characterized (49, 50), their role in DNA repair pathways has not been clearly established. Previous studies have suggested a role for RecN in homologous recombination, either via its influence on kinetics of strand-invasion by the RecA filament or by facilitating global chromosomal cohesion between replicated sisters (43, 45, 51, 52). Indeed, the potential relevance of RecN in recombination-based repair is underscored by its extensive conservation with other core recombination-associated proteins, such as RuvABC and RecA across bacteria (53).

In conjunction with molecular regulators, homology search appears to be an efficient and robust process; recent studies reveal that long-distance search and repair can typically be completed within a single generation time of bacterial growth (2, 3, 5). How RecA orchestrates this search, the dynamics of the nucleoprotein filament during this process, and the role for RecA regulators, such as RecN, in facilitating or modulating homology search, remain largely unknown. To elucidate the dynamics of homology search in vivo and investigate the role of RecN in this process, we use quantitative live-cell imaging to track RecA during DSB repair in the bacterium Caulobacter crescentus.

Here we present evidence for a search process involving a RecA nucleoprotein filament that translocates in a directional manner from the break site located at one cell pole to the opposite cell pole. The RecA filament simultaneously undergoes dynamic changes to its length, while traversing back-and-forth across the cell before finding its homologous partner and initiating homology pairing. This large-scale movement of RecA is contingent on RecN action. Such RecN-driven regulated and directional translocation of the RecA nucleoprotein filament points to a robust strategy of iterative search, thus enabling recombination between distant sites of homology.

Results

In Vivo Imaging of RecA Reveals Key Steps of Homologous Recombination.

To follow RecA-dependent search dynamics in vivo, we used the I-SceI system developed in C. crescentus (3, 33) that allows us to visualize DSB repair between distant homologous regions (SI Appendix, Fig. S1A). Briefly, we inserted a unique restriction site, I-SceI, near the replication origin, which is usually polarly localized in the cell. We induced a DSB by regulated expression of the endonuclease, I-SceI, that cleaves the site to generate a break. We further marked a region of the chromosome near the break-site using fluorescently-tagged MipZ, which associates with the origin region. DSB induction results in loss of localization of the MipZ marker close to the break (due to resection of DNA by the helicase-nuclease complex AddAB) (3, 33), serving as a reliable proxy to identify cells where only a single copy of the chromosome is cleaved (with only one MipZ marker now being visible). Restoration of the second marker is indicative of repair. Such complete repair events, as reported previously (3), were observed in 30% of all imaged cells during the course of a single experiment (with 22% cells losing both MipZ foci [suggesting two breaks], 18% cells facing no breaks, and 30% cells not having completed repair within the imaging window). Using this extensively characterized system (3, 33), we followed the process of homology search in two distinct cell types, both blocked for new rounds of replication initiation: predivisional cells with two completely replicated and segregated chromosomes (to follow homology search and subsequent recombination-mediated repair of distant homologous regions), and non-replicating swarmer cells with a single chromosome (to follow homology search, without the presence of the homologous repair template) (3, 33).

In both cell types, we monitored the dynamics of RecA using live-cell imaging of RecA-YFP. Given that endogenously tagged RecA perturbs the functioning of the protein (3–5, 45, 54) (SI Appendix, Fig. S1B), we used a well-described strategy in E. coli, where recA-YFP is expressed ectopically in a wild-type recA background (3–5, 45, 54). In this strain, as previously described in Caulobacter (3), recA is fluorescently-tagged at its C-terminus and expressed ectopically at low levels from a xylose-inducible promoter (on the chromosome) in a wild-type recA background. In our experimental set-up, RecA-YFP is expressed at 20 to 30% of wild-type RecA levels (SI Appendix, Fig. S1E). This allows us to avoid issues arising from overexpressing the tagged version of RecA (as reported in the case of E. coli) (41). This construct is found to be comparable to the wild-type strain in terms of function in generic DNA damage repair and recombination-mediated DSB repair (as assessed via CFU survival measurement on generic as well as DSB-inducing damage) (SI Appendix, Fig. S1 B–D).

Multiple additional observations support the validity of our system: 1) RecA-YFP or RecA-mCherry localization is contingent on the presence of DNA damage (Fig. 1A and SI Appendix, Fig. S4G). This is in contrast to RecA tags in E. coli that can localize polarly even in the absence of damage (4, 45, 54). 2) RecA localization persists only until damage repair. Upon DSB repair (reemergence of two MipZ foci), RecA does not localize in a discrete manner in the cell (Fig. 1A) (n = 100 cells). 3) RecA localizes at the site-specific break only. We find that RecA localization correlates with the spatial location of the break site within the cell. (SI Appendix, Fig. S2; representative from n = 100 cells in each example). 4) RecA-YFP can physically interact with wild-type RecA as well as a known interacting partner, RecN (SI Appendix, Figs. S1G and S3D). Furthermore, RecA-YFP alone (in a recA-deletion background) faithfully localizes to a DSB as foci, in response to break induction, and no localization is observed in the absence of damage. However, filament dynamics described in the following sections are contingent on the expression of the tagged construct in the presence of wild-type RecA, consistent with the likely formation of mixed filaments with both tagged and untagged RecA (SI Appendix, Figs. S1F, S2 A and E). 5) RecA and RecN localizations (see below) are only observed in the presence of DNA damage and they are always colocalized (SI Appendix, Figs. S3 A–C and S4G). 6) Readout of SOS response [using mitomycin-C (MMC) or Norfloxacin treatment, following previously described protocols in Caulobacter (3)] shows that cells carrying the fluorescently tagged RecA construct induce the response similar to wild-type conditions (SI Appendix, Fig. S1C).

Fig. 1.

Dynamics of RecA filament during homologous recombination. (A) Montage of key events of homologous recombination observed via live-cell imaging of both break/homology site (MipZ; red) as well as DSB dynamics (RecA; green). (Upper) Representative cell examples for each stage (t0 refers to the first time point of imaging); (Lower) quantitative estimation of duration of the key events described across multiple cells (n = 100 repair events). Single DSB induction (white asterisk) is marked by the disappearance of one of the two MipZ localizations. RecA nucleation is defined as the period between DSB induction and the initiation of RecA localization at the break site, followed by homology search, marked by the translocation of the RecA filament in the cell. Repair is marked by the loss of RecA filament after search, followed by the reappearance of two MipZ markers. The time line below shows the period of RecA nucleation and loading (after break induction), homology search and repair (n = 100). (Scale bars, 2 μm.) (B) Montage of RecA-YFP tracked over time (duration of a single pole-to-pole traversal event) after induction of a single DSB during homology search (white asterisk). Here, t0 is defined as the time point prior to the observation of a positive displacement of the RecA localization from the pole at which it first localized (DSB pole). (Scale bar, 2 μm.) (C) Representative RecA-YFP fluorescence profiles across (Left) one pole-to-pole traversal and (Right) multiple traversals.

As stated above, we find that Caulobacter RecA-YFP forms discrete structures (foci or elongated filaments) only in the presence of a DSB. In the present study, we use the term “filament” to describe diffraction-limited, extended RecA localizations (distinct from a spot or foci). Whether this includes bundles or a single filament of RecA polymers (4, 5) remains to be determined. In the absence of damage, fluorescence is diffuse throughout the cell (SI Appendix, Fig. S4G). We observed three distinct stages following a DSB near the cell pole: 1) nucleation and growth of RecA filament near the pole where MipZ localization, proximal to the origin (3, 33), is lost (indicating occurrence of a DSB); 2) homology search, where the RecA filament moves directionally across the length of the cell (Fig. 1 B and C) (expanded in the next sections); and 3) repair, associated with RecA dissociation followed by the reappearance of two distinct MipZ markers. An example of this process can be observed in Fig. 1A, which also provides the average duration of each stage described (n = 100 complete repair events; t0 refers to the first time point of imaging).

Growth and Directional Movement of RecA Filament during Homology Search.

We now characterize the dynamics of RecA during the homology search phase. In contrast to the static structures previously reported in E. coli (4, 5), we find that once RecA nucleates at the DSB site near the pole, it rapidly grows into an elongated filament, following which it dynamically moves across the length of the cell (Fig. 1 B and C, SI Appendix, Fig. S2 A and D, and Movie S1). We followed the time series of the centroid position x_c(t) of the RecA filament along the long-axis of the cell, relative to the cell length L (Fig. 2 A and B) (while analyzing traversals, t0 is defined as the time point prior to the observation of a positive displacement of the RecA localization from the pole at which it first localized) (SI Appendix, Supplementary Method includes details of quantitative analysis). From its initial position x_c(0), we found that the centroid progresses in a stochastic but biased manner, toward the opposite cell pole (Figs. 1B and C and 2B). Having reached the opposite cell pole, the RecA filament starts translocating in the reverse direction (Fig. 1C, Right, and SI Appendix, Fig. S2A). Defining a pole-to-pole translocation as a traversal, we find that the mean traversal time of the RecA filament is 6 ± 2 min, and typically undergoes multiple traversals (mean = 4) before homology pairing.

Fig. 2.

Directional translocation of the RecA filament during homology search. (A, Upper) Schematic showing measurement of centroid position of the RecA filament x_c (measured relative to cell length L) and the centroid steps Δx_c, (the algebraic difference between the centroid positions at two consecutive time points) and (Lower) schematic showing measurement of RecA filament lengths (l) (measured relative to cell length L) and length changes Δl (the algebraic difference between the filament lengths at two consecutive time points). (B) Centroid position of the RecA filament x_c versus time over a single traversal. Plot shows mean and SD from data collected over n = 99 traversals. Mean traversal time t = 6 ± 2 min is indicated. (C) Bar graph representing the probability distribution of the centroid steps Δx_c, (i.e., the algebraic difference between the centroid positions at two consecutive time points, 60 s apart), shows a bias toward positive values and a weight at Δx_c = 0 (error bars represent sample error). (Inset) Bar graph representing the probability distribution of x_c taken over a single traversal cycle (n = 795). All data shown are from predivisional cells. (D) Representative images of the types of RecA localizations observed during a wild-type traversal, with (images 1 and 2) showing puncta and (images 3 to 9) showing extended filamentous structures. Data shown are from predivisional cells. (Scale bar, 2 μm.) (E) Joint probability distribution of l and x_c taken over a single traversal in wild-type cells (n = 795). Data shown are from predivisional cells. (F) Probability distribution of the length changes Δl, (i.e., the algebraic difference between the filament lengths at two consecutive time points, 60 s apart) in wild-type cells (n = 347). Error bars represent sample error. (Inset) Density histogram of fraction lengths (l) of RecA filaments for wild-type cells (n = 795). Red dashed line is indicative of mean. Data shown are from predivisional cells.

On average, the centroid position is uniformly distributed across the cell length (Fig. 2 C, Inset). This indicates that there are no hot-spots within the cell that specifically regulate RecA filament dynamics. The distribution of the centroid steps Δx_c, the algebraic difference between the centroid positions at two consecutive time points (Fig. 2A), exhibits a bias toward positive values, indicating a systematic movement of the RecA filament in one direction. The nonzero weight of this distribution at Δx_c = 0 is an indication that the RecA filament often stalls (Fig. 2C).

Significantly, we observed that when the DNA break is made midcell (+780 kb from the origin) (3), the RecA filament nucleates near midcell, and moves toward either one of the cell poles (SI Appendix, Fig. S2B). Similarly, in non-replicating swarmer cells with a single copy of the chromosome, the RecA filament seems to exhibit interminable pole-to-pole traversals (Fig. 3F, SI Appendix, Fig. S2C, and Movie S2). These observations imply that the homology search dynamics proceeds independent of the homologous repair template.

Fig. 3.

RecN is essential for RecA filament movement in vivo. (A) Montage of RecA-YFP tracked over multiple pole-to-pole traversal events after induction of a single DSB (white asterisk) (Upper) in wild-type, (Lower) in cells lacking recN (for the same duration as wild-type cells above). Here, t0 is defined as the first time point when a RecA localization is observed. Examples shown from predivisional cells. (Scale bars, 2 μm.) (B) Representative RecA-YFP fluorescence profiles in a recN deletion background. (Upper) RecA is tracked for the expected duration of two pole-to-pole traversals; (Lower) over expected duration of multiple traversals. Data shown are from predivisional cells. (C) Centroid position of the RecA filament x_c versus time for cells lacking recN. Plot shows mean and SD from data collected over n = 50 cells. Data shown are from predivisional cells. (D) Probability distribution of the centroid steps Δx_c shows no bias (error bars represent sample error). (Inset) Probability distribution of x_c taken over all time points (n = 852), in cells lacking recN. Data shown are from predivisional cells. (E) Representative RecA-YFP fluorescence profile in a recN deletion background. DSB is induced in non-replicating swarmer cells with a single chromosome, thus lacking a homologous template for repair. (F) Probability distribution of x_c in wild-type (n = 911) and cells lacking recN (n = 1,039). DSB is induced in non-replicating swarmer cells with a single chromosome, thus lacking a homologous template for repair. (G) Probability distribution of the length changes Δl, (i.e., the algebraic difference between the filament lengths at two consecutive time points, 60 s apart) in wild-type (n = 347) and cells lacking recN (n = 750). Graph for wild-type cells is the same as in Fig. 2F. Error bars represent sample error. Data shown are from predivisional cells. (H) Scatter plot for l versus Δl in wild-type (n = 430) and recN-deleted cells (n = 750). Pearson correlation coefficient r = −0.466 (95% CI [−0.5372 to −0.3889]) for wild-type and r = −0.332 (95% CI [−0.3943 to −0.2669]) for recN deletion cells. Data shown are from predivisional cells.

Although there are two break ends upon DSB induction, we rarely observe two distinct RecA filaments. Thus, RecA appears to form a single filamentous structure spanning both break ends, which translocates across the cell during homology search. Importantly, the length of the RecA filament is typically a small fraction of the cell length and does not extend over the entire cell; the filament mean fractional length is l = 0.2 ± 0.1 (Fig. 2F, Inset), which corresponds to 1.0 ± 0.5 μm for cell lengths that range between 5 μm and 6 μm.

We find that directional movement of RecA during a homology search is accompanied by dynamic changes in the RecA filament length over time. For example, we observe various types of RecA structures in the cell during a traversal, ranging from a focus to extended or even bent filamentous structures (Fig. 2D). We find a weak correlation between the length of the filament and intensity of RecA (SI Appendix, Fig. S4F). However, we do not observe a significant correlation between the length of the RecA filament and its centroid position in the cell (Fig. 2E), suggesting that movement of the RecA filament is accompanied by stochastic variation of its length. In support, we note that the distribution of Δl, the algebraic difference between the filament lengths at two consecutive time points, shows a large range during a translocation (Fig. 2F).

RecN Is Essential for Movement of the RecA Filament In Vivo.

We now inquire into the molecular basis of the dynamics of the RecA filament during homology search and for holding break ends together during filament translocation. An essential and highly conserved member of the bacterial recombination pathway is the SMC-like protein, RecN (44, 46). However, it still remains unclear as to how RecN influences RecA-mediated homologous recombination in vivo. Given the universal role of SMC proteins in organizing DNA (48) (for example, via DNA bridging or DNA loop extrusion), we ask whether RecN contributes to RecA dynamics and organization of break ends specifically during homology search.

Consistent with reports in other bacterial systems (44, 46, 47), we find that RecN is essential for successful DSB repair in Caulobacter. In the absence of recN, we record no detectable repair events, together with an increased sensitivity to DSB-inducing agents (SI Appendix, Figs. S3 E and F and S4H). We use the bacterial two-hybrid system to show that Caulobacter RecA and RecN physically interact with each other (SI Appendix, Fig. S3D). We also find that the recruitment of RecN to damage sites is RecA-dependent (and not only on SOS-dependent expression) (SI Appendix, Fig. S3 A and B). Furthermore, RecA and RecN are colocalized at a DSB site, with no detectable instances of RecN localizing away from the RecA structure. However, we did observe examples of RecN localizing toward one end of an elongated RecA filament (SI Appendix, Fig. S3C).

Given these observations, we investigate how RecN impacts RecA dynamics and homology search. For this, we first delete recN and follow the dynamics of RecA during the search process via live-cell imaging. As in the case of wild-type, we still observe a single RecA localization in recN cells, suggesting that RecN does not contribute to holding break ends together. Furthermore, RecA filament length in recN-deleted cells is similar to wild-type and also does not span the entire length of the cell (mean fractional length is l = 0.2 ± 0.06, which corresponds to 1.2 ± 0.5 μm for cell lengths that range between 7 μm and 9 μm). Instead, we find that upon recN deletion, the RecA filament ceases to translocate in predivisional cells (Fig. 3 A–D and Movie S3) (here, t0 is defined as the first time point when a RecA localization is observed). Unlike wild-type, the RecA filament remains rooted to the pole where the break occurs (Fig. 3 B–D, Inset) and does not undergo pole-to-pole traversal over the entire course of imaging, up to 100 min (Fig. 3C). Concomitantly, no RecA dissociation is observed, consistent with the idea that recombination between distant homologous regions involves RecN-driven RecA movement. In our experiments, all repair events are associated with the localization and mobility of RecA filaments (and search across the cell). Lack of search (as seen with the absence of RecN) results in zero repair events (SI Appendix, Fig. S4H).

We find a similar effect on the RecA filament translocation in the absence of recN, in non-replicating swarmer cells with only a single copy of the chromosome, in support of our assertion that RecN-driven RecA movement for homology search occurs independent of a homologous template for repair (Fig. 3 E and F and Movie S4). Finally, these perturbations to RecA dynamics are observed only in cells lacking the repair-specific SMC-like protein (RecN) and not the SMC protein (Smc) involved in chromosome organization and segregation in non-damage conditions (48) (SI Appendix, Fig. S3 F and G).

To study which aspects of RecA activity are perturbed in the absence of recN, we assess the effects of recN deletion at different stages of the RecA dynamics. In the nucleation stage, we measure the time taken from MipZ disappearance to the appearance of a RecA nucleation event at a DSB site, and find it comparable to wild-type cells (SI Appendix, Fig. S4A). In the growth stage following nucleation (and prior to search), we find that the initial rates of RecA loading are unaltered in recN-deleted cells (SI Appendix, Fig. S4 D and E). In line with this, we observe that the frequency of RecA nucleation events across recN cells is comparable to wild-type, even under generic MMC damage (SI Appendix, Fig. S4G).

We next examine whether the absence of RecA filament movement in recN-deleted cells is a consequence of RecA filament instability. Previous studies have shown that cells lacking recA or with mutants of RecA that are unable to stably associate with a DSB, undergo “rec-less” degradation, resulting in excessive loss of DNA (due to unregulated resection) around a break site (30, 55). Hence, we measured DNA loss in cells with a single DSB (+780 kb from the origin in non-replicating swarmer cells) in wild-type, recA, and recN backgrounds, via a deep-sequencing assay (33, 56). One hour after damage induction in the wild-type background, lack of detectable sequencing reads around the break site is indicative of the proportion of cells that have undergone a DSB. The sequencing profile is asymmetric around the DSB and progressively increased in both directions, until no read loss is observed (matching that of the control). This pattern is consistent with previously published observations (33, 56). Importantly, cells lacking recN have resection profiles that overlap with that of wild-type (SI Appendix, Fig. S4 B and C), while cells lacking recA show extensive resection around the break site. In addition, recN-deleted cells do not have perturbed SOS induction (SI Appendix, Fig. S4I). Together, these observations suggest that RecA filament association with a DSB is not affected in the absence of recN, although its movement during homology search is stalled.

In wild-type cells, RecA translocation is accompanied by changes in the filament length (Fig. 2 D–F). We hence ask whether such RecA length variation is also perturbed in cells lacking recN. In comparison to wild-type, we find that, in recN-deleted cells, the filament lengths tend to be less variant between two consecutive imaging frames (across a range of filament lengths) during the time period when RecA translocation is anticipated (Fig. 3 G and H). Thus, taken together, these results implicate a role for RecN in driving both RecA mobility as well as filament length variations during homology search.

RecN ATPase Cycle Regulates RecA Filament Dynamics during Homology Search.

While the molecular mechanism of filament translocation warrants future investigations (Discussion), the following additional observation underscores a central role for RecN in modulating RecA filament lengths as well as mobility: A fundamental characteristic of SMC family proteins is their ATPase cycle, that enables them to influence the organization of the substrates they are bound to as well as translocate across large distances within the cell (49, 57). We hence ask how the RecN ATPase cycle regulates RecA filament dynamics, thus facilitating recombination between distant homologous regions. For this, we generate two mutants of RecN in the conserved WalkerA and WalkerB motif that are likely to result in specific perturbation to either ATP binding or ATP hydrolysis properties of RecN (49, 58): recN^D472A (ATP-binding impaired) and recN^K35A/E473Q (ATP-hydrolysis impaired) (59). We find that both mutants are compromised in DNA damage repair, similar to the recN deletion (Fig. 4C). We next follow the dynamics of RecA during homology search after induction of a DSB in cells carrying either the recN^D472A or recN^K35A/E473Q mutations. We find that these mutants also result in static RecA filaments that are unable to translocate across the cell for homology search (Fig. 4 A, B, and D–F and Movies S5 and S6). While translocation of RecA is similarly impaired in both mutants, we notice that RecA filament lengths are differently affected. In the case of recN^D472A, RecA filament lengths are comparable to wild-type, while they are significantly shorter in the putative RecN ATP hydrolysis mutant (Fig. 4G).

Fig. 4.

RecN ATPase cycle regulates RecA filament dynamics. (A) Montage of RecA-YFP tracked over time after induction of a single DSB (white asterisk) for putative RecN ATP hydrolysis mutant (recN^K35A/E473Q). (Scale bar, 2 μm.) (B) Representative montage of RecA-YFP tracked over time after induction of a single DSB (white asterisk) for putative RecN ATP binding mutant (recN^D472A). (Scale bar, 2 μm.) (C) Sensitivity of recN^D472A and recN^K35A/E473Q to DSB-inducing agent norfloxacin is evaluated (in comparison to wild-type, ΔrecN, and ΔrecA cells, from SI Appendix, Fig. S3E). Representative image from at least three independent repeats. (D, Upper) Representative RecA-YFP fluorescence profile across the length of the cell in a recN^K35A/E473Q background. (Lower) Representative RecA-YFP fluorescence profile across the length of the cell in a recN^D472A background. (E) Position of centroid of RecA filament normalized to length of the cell (x_c) is plotted for recN^K35A/E473Q. Mean and SD is shown (n = 50). (F) As in E for recN^D472A. (G) Density histogram of RecA filament lengths (l) across all time points of imaging for wild-type, recN^D472A, and recN^K35A/E473Q (n = 229). Distribution of l is found to be significantly different between wild-type and recN^K35A/E473Q (P = 0.0005, 95% CI [0.008051, 0.02846], Welch’s t test). Distribution of l is comparable between wild-type and recN^D472A (P = 0.1535, 95% CI [−0.003326, 0.02108]). All data shown are from predivisional cells.

Taken together, the differential impact of putative RecN ATPase mutants on RecA filament lengths (both resulting in abrogation of mobility), leads us to suggest that RecN, via its ATPase cycle, likely modulates RecA nucleoprotein filament lengths and drives directional translocation of the filament during homology search.

Discussion

Using quantitative live-cell imaging, we have uncovered a new mode of RecN-dependent RecA action in long-distance homology search during recombination, a fundamental process for the maintenance of life. We find that following a site-specific DSB, RecA nucleates at the DSB site and grows into a nucleoprotein filament, which moves in a stochastic but directional manner along the length of the cell. Filament translocation is accompanied by variation in RecA filament length. The RecA filament makes multiple traversals across the cell before the homologous pair is found, followed by repair. Importantly, RecN, via its ATPase cycle, is essential for this search process and the subsequent repair by homologous recombination.

Together, based on previous findings and our present study, the following series of events during the early steps of homologous recombination following a DSB emerges (SI Appendix, Fig. S5): 1) DNA end resection by the helicase-nuclease complex AddAB (or RecBCD) to reveal ssDNA ends (33, 56); 2) RecA loading (that can be facilitated by the helicase-nuclease complex itself) (60) on ssDNA, where it forms a nucleoprotein filament and/or bundle (4, 5); 3) RecN recruitment to the DSB by RecA (47); and 4) RecN-dependent RecA filament translocation for long-distance homology search. Although there are two break ends, both of which must search for and find the distant DNA homology if faithful recombination repair is to succeed, we observe only a single contiguous RecA filament. Furthermore, RecN does not appear to contribute to holding these ends together. Instead, our observations position RecN early in the recombination pathway, specifically in enabling RecA filament mobility during the homology search, with RecN influencing RecA dynamics even in the absence of a homologous repair template. This is similar to other SMC-like protein complexes, such as the eukaryotic Rad50-Mre11 (61) or bacterial SbcCD (62) (conserved only in gamma-proteobacteria) (53), which also play a key role in DSB repair, prior to strand exchange.

The RecN-driven RecA dynamics and the multiple traversals before identification of the homologous repair template, suggests a new mode of iterative, long-distance search that allows for a regulated genome-wide sampling for homology. Earlier work has suggested that RecA engages in a one-dimensional sliding movement and intersegment hopping, while regions within the RecA filament engage in microhomology sampling during search (31, 63, 64). Our observations reveal that in addition to the initial waiting time between nucleation to the start of traversals, the translocating RecA filament makes pauses at random intervals. It is possible that RecN-dependent RecA dynamics reflects such microhomology sampling activity prior to homology pairing, with RecA molecules engaging in microhomology sampling both initially in the vicinity of the damage site (especially in the postreplicative cohesion period) and during these pause intervals during translocation (38, 63). Significantly, RecN-mediated multiple pole-to-pole traversals by RecA builds in a kind of reset at a fixed spatial location, a strategy that can lead to optimal search (65), thus facilitating repair well within a single generation time of bacterial growth.

How does RecN influence RecA mobility to drive directional translocation? While unanswered presently, an immediate future goal is to unravel the mechanistic basis underlying RecN-regulated dynamics of the RecA nucleoprotein filament revealed here. This would require a quantitative characterization of the complete RecA-RecN ATPase cycle, its interaction with the mechanochemistry of the DNA substrate, as well as detailed dissection of RecA filament architecture, RecA-RecN stoichiometry, and turnover. In vivo studies in E. coli have suggested that the DSB-associated RecA structures could either be a bundle (4) or a single (5) polymer of RecA molecules. It would be informative to determine the detailed composition and dynamics of the RecA filament during translocation and how these are impacted by RecN. For example, we find that RecN colocalizes with RecA, with instances where RecN is localized at the base of a RecA filament. One possibility is that localized action of RecN at one end of the RecA filament contributes to the directionality of the movement observed.

In this regard, the reported properties of RecN are also important to consider. Studies have shown that RecN can bridge ssDNA and double-stranded DNA (52), and increase global chromosome cohesion in response to DNA damage (45). Thus, in light of RecN being an SMC-like protein (48, 51, 59) and its DNA binding properties (51, 52), it is tempting to invoke a possibility of motor-driven movement [such as extrusion (66) or zipping (49)] of RecA and the underlying DNA substrate. Such motor-driven movement could result in variation in RecA filament length during translocation (such as bending or twisting of the filament). Alternatively, in vitro studies have shown that RecN can stimulate RecA strand invasion (43, 52). Separately, RecA ATPase rates (and likely turnover) are stimulated fourfold upon strand invasion, potentially allowing the RecA filament to release from heterologous pairing during microhomology sampling (38). Thus, RecN could act more locally on the RecA filament via influencing the assembly–disassembly rates of monomers within the filament and at filament ends. Indeed, these are not exclusive scenarios and detailed molecular characterization of the RecA-RecN dynamics will enable mechanistic understanding into this process. We note that a caveat and current limitation in the in vivo imaging of RecA is that fluorescently labeling RecA does compromise its function; however, our observations suggest that the RecA dynamics observed are arising from the formation of mixed filaments (between RecA-YFP and untagged-RecA) and are consistent with the fusion reported here being indicative of damage-associated activity.

In conclusion, our work reveals a central role for regulation of RecA filament dynamics by RecN during long-distance homology searches. While previous studies have established RecN as a key player in the homologous recombination pathway, its precise contributions to the process has remained unanswered. Here we have uncovered three key elements of the RecN-driven homology search in vivo: RecA filament mobility, associated changes in filament length, and the ability to undertake multiple pole-to-pole traversals. It is possible that there is some diversity in RecA-dependent long-range homology search dynamics across bacterial systems, to contend with the varying cell shapes, chromosome replication dynamics (multifork vs. replication only once per cell cycle), and chromosome organization (2–5), thus resulting in static (4, 5) or mobile RecA structures (present study). However, given that RecA and RecN are among the most conserved proteins of the recombination pathway in bacterial genomes and its counterparts are present across all domains of life, we believe that the dynamics revealed here occur in other systems as well.

Materials and Methods

Bacterial Strains and Growth Conditions.

Strains, plasmids, and oligos used in this study are described in SI Appendix, Tables S1–S3, respectively. Chromosomal modifications, such as integration of fluorescent markers or deletion of genes, were performed using the two-step recombination method or with vectors described previously (67). Transductions were carried out using ФCr30. Unless otherwise stated, Caulobacter cultures were grown at 30 °C in peptone yeast extract and supplemented with antibiotics at appropriate concentrations. When growing strains with dnaA under an isopropyl-β-d-thiogalactopyranoside (IPTG)-inducible promoter, liquid media was supplemented with 0.5 mM IPTG and solid media with 1 mM IPTG. For synchronization experiments, Caulobacter cultures were grown until midlog phase and synchronization protocols were followed as previously described (3). Protocols used for isolating non-replicating predivisional cells and nonreplicating swarmer cells (3, 33) are briefly described here. For isolation of non-replicating swarmer cells, cultures were grown until midlog phase in media supplemented with IPTG. Cells were then depleted for DnaA by washing off the inducer and were grown in the absence of IPTG for one generation. This was followed by synchronization and isolation of swarmer cells. These cells were resuspended in media without IPTG and used for DSB induction experiments. For isolation of non-replicating predivisional cells, cultures were grown until midlog phase in media supplemented with IPTG. Cells were then synchronized and swarmer cells isolated. Swarmers were then released into media lacking IPTG. Cells were then depleted for DnaA by washing off the inducer and grown in the absence of IPTG for one generation. Cell division was blocked via addition of 35 μg/mL cephalexin and these non-replicating predivisional cells were further used for DSB induction experiments. In all cases, recA-YFP was induced using 0.03% xylose 90 min prior to imaging and subsequently maintained in the agarose pad as well during the entire course of imaging.

Fluorescence Microscopy.

Time course imaging was performed on 1% agarose pads (Invitrogen ultrapure). For time-lapse imaging, cells were grown on 1.5% GTG agarose pads (low melting) prepared in peptone yeast extract and supplemented with vanillate, xylose, or cephalexin at appropriate concentrations and imaged using glass-bottom Petri dishes. Imaging was carried out using a wide-field epifluorescence microscope (Eclipse Ti-2E, Nikon) with a 63× oil immersion objective (plan apochromat objective with NA 1.41), illumination from a pE4000 light source (CoolLED), Hamamatsu Orca Flash 4.0 camera, and a motorized XY stage. During time-lapse imaging, focus was maintained using an infrared-based Perfect Focusing System (Nikon). Image acquisitions were done using NIS-elements software (v5.1) and images were acquired every 30 s, 1 min, or 5 min (as indicated in the main text and respective figure legends). For most experiments with RecA, exposure time used for excitation at 490 nm was 400 ms and for 550 nm was 500 ms. For imaging MipZ, exposure time used for excitation at 550 nm was 100 ms. For imaging YFP expressed from the sidA promoter, exposure time used for excitation at 490 nm was 500 ms. In all images, scale bars = 2 μm. Procedures for image analysis and for extracting key features of RecA localization are described in SI Appendix, Supplementary Method.

Whole-Genome Sequencing Assay to Measure DNA Loss around the DSB Site.

Whole-genome sequencing was implemented as previously described (33). Briefly, non-replicating swarmer cells with a DSB site at +780 kb were isolated and released in a media with 0.5 mM vanillate to induce DSBs. Samples were collected at 0 h (control) and 1 h after DSB induction. Genomic DNA was isolated using a DNAeasy blood and tissue kit (Qiagen). Whole-genome Illumina sequencing was carried out on the isolated DNA (Next Generation Sequencing facility, National Center for Biological Sciences, India). Data were further analyzed using the following protocol. First, indexing with the reference genome (4.01 Mbp) (National Center for Biotechnology Information reference sequence: NC-011916.1) was done using BWA (68). Reads with raw read quality 20 were aligned using “BWAaln –q.” SAMTOOLS (v0.1.19-96b5f2294a) (68) was used to filter out the multiply mapped reads. Finally, with BEDTOOLS (69) the read count per bin was calculated using the .bed files containing bin positions. Caulobacter genome data obtained after Illumina sequencing, HisEq. 2500 Illumina short reads (50 bp), was divided into 1-kb bins. Read counts per bin were normalized to the total reads acquired for that sample. The ratio of normalized reads after DSB (1 h) to normalized reads before DSB (0 h) was plotted to visualize the read enrichment profile obtained across the genome after induction of a DSB. The graph generated was processed further with the Lowess smoothing function in MatLab.

Bacterial Two-Hybrid Assay.

The bacterial two-hybrid assay was implemented as previously described (33). Briefly, to investigate physical interaction between a pair of proteins, their respective genes were fused to 3′ end of T25 or T18 fragments in pKT or pUT vectors. These vectors were cotransformed into E. coli BTH101. Cotransformed cells were grown to saturation in M63 media with maltose and IPTG and 5 μL of this culture was spotted on MacConkey agar plates (40 g/L) with maltose, IPTG, and appropriate antibiotics. Plates were incubated at 30 °C for 2 to 3 days.

Western Blotting.

Predivisional cells were released in fresh PYE supplemented with 2 µM vanillate to induce DSBs. A 6-mL culture of OD₆₀₀ 0.1 cells was collected at 0 h and 1.5 h after DSB induction. Pellets were resuspended in 100 µL of 1× sample buffer (For 4× sample buffer: 500 mM Tris⋅HCl [pH 6.8], 8% SDS, 40% glycerol, bromophenol blue) with 4% (vol/vol) β-mercaptoethanol. Cell lysis was induced by heating at 95 °C with intermittent vortexing. Equal volumes of samples were loaded on 12% polyacrylamide gels. Wet transfer was used to blot proteins on a PVDF membrane. Post-transfer, 5% nonfat milk in TBS-T was used for blocking. E. coli RecA (1:5,000; Abcam) and RpoA (1:5,000; Biolegend) antibodies reconstituted in 3% nonfat milk in TBS-T were used for probing. Appropriate HRP-conjugated secondary antibodies (1:5,000) were reconstituted in 0.3% nonfat milk in TBS-T and used. Blots were developed using SuperSignal West PICO PLUS and quantification was carried out using ImageJ. RpoA was used as a loading control.

Data, Materials, and Software Availability

All study data are included in the main text and supporting information. Sequencing data are deposited on NCBI with accession number PRJNA741456. All analysis codes can be found at https://github.com/badrinarayanan-lab/Chimthanawala-et-al-2022 (70).