Bacillus subtilis RadA/Sms-Mediated Nascent Lagging-Strand Unwinding at Stalled or Reversed Forks Is a Two-Step Process: RadA/Sms Assists RecA Nucleation, and RecA Loads RadA/Sms

Replication fork rescue requires Bacillus subtilis RecA, its negative (SsbA) and positive (RecO) mediators, and fork-processing (RadA/Sms). To understand how they work to promote fork remodeling, reconstituted branched replication intermediates were used. We show that RadA/Sms (or its variant, RadA/Sms C13A) binds to the 5′-tail of a reversed fork with longer nascent lagging-strand and unwinds it in the 5′→3′ direction, but RecA and its mediators limit unwinding. RadA/Sms cannot unwind a reversed fork with a longer nascent leading-strand, or a gapped stalled fork, but RecA interacts with and activates unwinding. Here, the molecular mechanism by which RadA/Sms, in concert with RecA, in a two-step reaction, unwinds the nascent lagging-strand of reversed or stalled forks is unveiled. First, RadA/Sms, as a mediator, contributes to SsbA displacement from the forks and nucleates RecA onto single-stranded DNA. Then, RecA, as a loader, interacts with and recruits RadA/Sms onto the nascent lagging strand of these DNA substrates to unwind them. Within this process, RecA limits RadA/Sms self-assembly to control fork processing, and RadA/Sms prevents RecA from provoking unnecessary recombination.


Introduction
Replication fork progression is frequently impeded by physical barriers (damaged bases, DNA distortions, protein barriers), shortage of DNA precursors, or collision with transcription elongation complexes or R-loops [1][2][3][4]. This leads to replication stress, which is a major contributor to genomic instability [2,3]. In some cases, the lesion/obstacle is simply skipped. In other cases, homologous recombination (HR) functions may catalyze reversion of the stalled replication fork leading to a structure that resembles a Holliday junction (HJ), a process also called fork regression [2,3,[5][6][7]. In both, the aim is to generate an intermediate for the resumption of DNA replication [2,3,[5][6][7].
The model bacteria Escherichia coli and Bacillus subtilis, which diverged >1.5 billion years ago and show a genetic distance larger than that between humans and plants, use similar pathways to overcome a replication stress, albeit with different hierarchal order. (Unless stated otherwise, indicated genes and products were of B. subtilis origin.) In UVirradiated E. coli cells, DNA synthesis drops within a few minutes, leading to a stable average number of replisomes [8] but a variable average number of DNA polymerase (DNAP) III Eco spots [8,9]. Here, RecA Eco forms foci at locations distal from replisomes iñ 65% of the cases [8]. Similarly, RecO Eco , crucial for RecA loading, rarely coincide with DNAP III Eco [10]. It is likely that in E. coli, when a replication fork encounters a physical impediment, DNAP III Eco skips the barrier to resume synthesis downstream. Next, RecA Eco mainly assembles at the gap left behind and facilitates gap filling by different DNA damage tolerance sub-pathways [11][12][13]. However, up to~35% of total RecA Eco foci colocalize with DNAP III Eco components, perhaps when lesion skipping is impeded by trafficking problems [4,14]. Here, DNAP III Eco disassembles, and the stalled fork is pushed backwards synthesized leading-strand and no-synthesis in the lagging-strand (3 -fork DNA) and unwind the template lagging-strand with similar efficiency [34]. However, they cannot unwind duplex or D-loop DNA substrates since they lack a 5 -tail [34]. It is likely that the 5 -tail of the template lagging-strand penetrates the central channel of the hexameric ring to unwind the 3 -fork DNA by moving in the 5 →3 direction [32,34]. However, a DNA helicase activity for RadA Eco has not been described [43]. It has been proposed that RadA Eco speeds up the initial step of RecA Eco -mediated DNA strand exchange, as RecA mediators do [43,44].
RadA/Sms cannot unwind a replicating fork with a fully synthesized lagging-strand and no-synthesis in the leading-strand (5 -fork DNA). In the presence of RecA, wt RadA/Sms can unwind the nascent lagging-strand of 5 -fork DNA, but RadA/Sms C13A, which fails to interact with RecA, cannot [32,34,45]. In other words, RecA interacts with and loads RadA/Sms onto the nascent lagging-strand of 5 -fork DNA, and RadA/Sms unwinds it by a poorly understood mechanism [45]. However, many molecular details of the RecA activating mechanism for RadA/Sms unwinding, and of the role of RecA mediators on RadA/Sms substrate preference, are still unsolved. Moreover, very little is known of the RadA/Sms and RecA interplay in other bacteria.
In this study, we report findings that contribute to understanding the interplay of B. subtilis RadA/Sms and RecA in the presence of negative (SsbA) and positive (RecO) mediators. To unravel the role of these proteins, synthetic DNA substrates mimicking stalled replication forks with a small gap in the template leading-(forked-Lead) or laggingstrand (forked-Lag), or reversed forks with a nascent leading-(HJ-Lead) or lagging-strand tail (HJ-Lag) longer that its complementary strand, were constructed (Supplementary Information (SI) Figure S1). We show that RadA/Sms assembles on the 5 -tail of HJ-Lag DNA and unwinds it, with RecA and its mediators (SsbA or RecO) limiting the unwinding reaction. A different outcome was observed when the HJ-Lead, forked-Lead, or forked-Lag DNA substrates were used. Here, RadA/Sms cannot unwind the substrate, and an interaction with RecA is necessary for unwinding. In the presence of SsbA, RadA/Sms, working as a specific mediator, antagonizes the negative effect exerted by SsbA on RecA nucleation onto the ssDNA 3 -tail of the HJ-Lead or ssDNA gap of forked-Lead or forked-Lag DNA. Here, RadA/Sms interacts with and inhibits the ATPase of RecA. Finally, RecA, as a specific loader, activates RadA/Sms to unwind their nascent lagging-strand, yielding (directly or indirectly) a 3 -fork product. This mechanism of replication fork restoration has two main advantages over canonical replication fork remodeling: a 3'-fork DNA intermediate cannot be cleaved by the RuvAB-RecU resolvasome and is a substrate suitable for PriA assembly [46].

RadA/Sms Partially Counters the Negative Effect of SsbA on RecA Nucleation onto cssDNA
Previously, it has been shown that: (i) SsbA binds ssDNA with~25-fold and >500-fold higher affinity than RadA/Sms·ATP and RecA·ATP, respectively [42,47]; (ii) SsbA blocks (by 20-fold, p < 0.01) the ATPase activity of RecA, competing with its nucleation and/or inhibiting its polymerization onto SsbA-coated ssDNA ( Figure 1B, line 1 vs. 4, Table S1) [28,29]; (iii) RadA/Sms and RadA/Sms C13A hydrolyze ATP with similar efficiency in the absence of cssDNA (Table S1) [34]; (iv) RadA/Sms C13A-mediated ATP hydrolysis is stimulated by 5-fold (p < 0.01) in the presence of cssDNA, but that by wt RadA/Sms is not ( Figure 1A,B, line 2, Table S1) [34]; and (v) RadA/Sms interacts with and inhibits the ssDNA-dependent ATPase activity of RecA, but RadA/Sms C13A does not ( Figure 1A,B, line 1 vs. 3, Table S1) [34,42]. In all reactions, the circular 3199-nt ssDNA (cssDNA, 10 μM in nt) was incubated with the indicated proteins and concentrations (RecA (800 nM), RadA/Sms or its mutant variants (30 nM), SsbA (150 nM), and RecO (100 nM)) in the order detailed, in buffer A containing 5 mM ATP and the ATP regeneration system, and the ATPase activity was measured for 30 min. The arrow denotes that the ssDNA was incubated (5 min at 37 °C) with the preceding protein(s) (denoted between parentheses) to preform cssDNA-protein complex(es). ATP hydrolysis measurement was started after all proteins were added. All reactions were repeated three or more times with similar results, and a representative graph is shown here. The mean ATPase activity Vmax values ± SEM, calculated from the slope of the curves, can be found in Table S1.
When RecA (1 RecA monomer/12.5-nt), SsbA, and RadA/Sms were incubated with cssDNA, the maximum efficiency of ATP hydrolysis was increased by ~10-fold (p < 0.01) when compared to the one observed by the RecA-cssDNA-SsbA complex ( Figure 1A, line 1 vs. 4 and 6, Table S1). The maximal rate of ATP hydrolysis, however, was still decreased by ~2-fold (p < 0.01) when compared with RecA alone ( Figure 1A, line 1 vs. 6, Table S1). It is likely that RadA/Sms partially counteracts the negative effect of SsbA on RecA ATPase activity. This effect cannot be attributed to a protein-protein interaction, because neither RadA/Sms (SI Appendix S1 and Figure S2A,B) nor RecA [47] physically interacts with SsbA. In all reactions, the circular 3199-nt ssDNA (cssDNA, 10 µM in nt) was incubated with the indicated proteins and concentrations (RecA (800 nM), RadA/Sms or its mutant variants (30 nM), SsbA (150 nM), and RecO (100 nM)) in the order detailed, in buffer A containing 5 mM ATP and the ATP regeneration system, and the ATPase activity was measured for 30 min. The arrow denotes that the ssDNA was incubated (5 min at 37 • C) with the preceding protein(s) (denoted between parentheses) to preform cssDNA-protein complex(es). ATP hydrolysis measurement was started after all proteins were added. All reactions were repeated three or more times with similar results, and a representative graph is shown here. The mean ATPase activity V max values ± SEM, calculated from the slope of the curves, can be found in Table S1.
When RecA (1 RecA monomer/12.5-nt), SsbA, and RadA/Sms were incubated with cssDNA, the maximum efficiency of ATP hydrolysis was increased by~10-fold (p < 0.01) when compared to the one observed by the RecA-cssDNA-SsbA complex ( Figure 1A, line 1 vs. 4 and 6, Table S1). The maximal rate of ATP hydrolysis, however, was still decreased by~2-fold (p < 0.01) when compared with RecA alone ( Figure 1A, line 1 vs. 6, Table S1). It is likely that RadA/Sms partially counteracts the negative effect of SsbA on RecA ATPase activity. This effect cannot be attributed to a protein-protein interaction, because neither RadA/Sms (SI Appendix S1 and Figure S2A,B) nor RecA [47] physically interacts with SsbA.
To further evaluate if RadA/Sms can displace SsbA from ssDNA, the wt protein was replaced by RadA/Sms C13A, since: (i) its ATP hydrolysis is stimulated in the presence of cssDNA, and (ii) it fails to interact with RecA and to inhibit is ATPase activity [34]. SsbA reduced by~2.5-fold the ATPase activity of RadA/Sms C13A (p < 0.01) ( Figure 1B, line 2 vs. 5, Table S1). The maximal efficiency of ATP hydrolysis in the presence of RadA/Sms C13A, RecA, SsbA, and cssDNA was significantly higher (~15-fold, p < 0.01) when compared with the maximal rate of ATP hydrolysis by the RecA-cssDNA-SsbA complex ( Figure 1B, line 6 vs. 4, Table S1). It is likely that the wt and RadA/Sms C13A enzymes can partially counter the negative effect of SsbA on the maximal efficiency of ATP hydrolysis by RecA. Here, the ssDNA-independent ATPase activity of RadA/Sms C13A may mask the analysis. Previously, it has been shown that: (i) RecO cannot hydrolyze ATP [30], and (ii) RecO (1 RecO monomer/50-nt) partially displaces SsbA and facilitates RecA nucleation and filament growth onto ssDNA coated by saturating SsbA concentrations (1 SsbA tetramer/ 33-nt) [29,30]. Thus, we analyzed whether SsbA and RecO counteract the negative effect that RadA/Sms exerts on the ATPase of RecA.
First, we confirmed that both the ssDNA-dependent or ssDNA-independent ATPase activity of wt RadA/Sms or RadA/Sms C13A was not affected by the addition of a circa saturating RecO concentration (1 RecO monomer/100-nt) ( Figure S3A). To avoid the noise introduced by the ssDNA-independent ATPase activity of wt RadA/Sms, it was replaced by the ATPase defective RadA/Sms K104R or RadA/Sms K255R mutant variants that still physically interact with RecA [34]. These mutant variants (1 hexamer/330-nt) also blocked the ATPase activity of RecA (p < 0.01) ( Figure 1C,D, line 1 vs. 2, Table S1). This inhibition is protein specific, because the ATPase activity of an unrelated ssDNA-dependent ATPase (PcrA) was not affected by wt RadA/Sms, RadA/Sms K104R, or RadA/Sms K255R (SI Appendix S2, Figure S3B).
From these data altogether we conclude that: (i) RadA/Sms-RecA, RadA/Sms K104R-RecA or RadA/Sms K255R-RecA complexes limit the ATPase of RecA and indirectly RecA dynamics; (ii) the presence of the two-component mediator (SsbA and RecO) cannot counteract RadA/Sms K104R or RadA/Sms K255R (Table S1); (iii) the interplay between RecA with RadA/Sms or SsbA-RecO may be mutually exclusive, because RecA nucleated on the preformed SsbA-cssDNA-RecO complexes is only partially sensitive to RadA/Sms K104Ror RadA/Sms K255R-mediated inhibition ( Figure 1C,D, line 6); and (iv) RadA/Sms interacts with and loads RecA on the SsbA-cssDNA complexes, but inhibits RecA redistribution on cssDNA (Table S1).
To further test whether RadA/Sms or its mutant variants (RadA/Sms C13A (unable to interact with RecA) or RadA/Sms K104A (unable to hydrolyze ATP)) can affect the enzymatic activity of RecA, a three-strand recombination assay was used. As revealed in SI Appendix S3 ( Figure S4A,B), RecA in the ATP bound form (RecA·ATP) or RecA·dATP-mediated DSE was significantly reduced and inhibited (p < 0.01) in the presence of~6 and 12 RadA/Sms, RadA/Sms C13A, or RadA/Sms K104A hexamers/cssDNA molecule, respectively (Supplementary Figure S4A,B). Alternatively, RadA/Sms bound to the DNA substrates reverses RecA·ATP-mediated DSE. Independently of the time point at which RadA/Sms is added, the reaction neither progressed further nor is reversed when compared to the absence of RadA/Sms (Supplementary Figure S4C). It is likely that six or more RadA/Sms, RadA/Sms C13A, or RadA/Sms K104A hexamers/cssDNA molecule limit RecA-driven DSE rather than stimulating the backwards reaction toward the original substrates. An uncharacterized function(s) should be necessary to overcome RadA/Smsmediated limitation of DSE, because RadA/Sms is crucial for strand assimilation during natural chromosomal transformation [32,34]. RecA forms foci that colocalize with the stalled replisome in >90% of cells after UVirradiation [19,22]. RadA/Sms forms dynamic foci that mostly colocalize with the DNA bulk in wt cells, but it forms static foci in the ∆recG context, where branched intermediates accumulate [38,40,41]. In vitro, SsbA bound to ssDNA blocks RecA nucleation on the SsbA-ssDNA complex, but RadA/Sms partially displaces SsbA from ssDNA ( Figure 1). Furthermore, RecA binds and protects branched intermediates (e.g., stalled or reversed fork) and interacts with and loads RadA/Sms at these substrates [34]. To test whether RadA/Sms in the presence of RecA and/or SsbA remodels or processes a stalled fork with a leading-strand gap that is reversed, a synthetic reversed fork with a nascent lagging-strand longer than the leading-strand (HJ-Lag) was constructed ( Figure S1).
The size site of RadA/Sms is unknown. To design the HJ-Lag substrate, we first analyzed the efficiency of unwinding of duplex DNAs with increasing 5 -tail lengths (SI Appendix S4). RadA/Sms poorly unwound a 15-nt long 5 -tail duplex, but it efficiently unwound a substrate with a 30-nt long 5 -tail ( Figure S5). It is likely that the RadA/Sms site size lay between these two values. The size sites of SsbA and RecA were also analyzed. SsbA binds ssDNA with a~30-nt average size (among the SSB 65 , SSB 35 , SSB 15 , and SSB 7 binding modes) [47,48]. RecA·ATP forms a discrete dynamic RecA nucleoprotein filament when~8 or more monomers bind ssDNA, with one monomer binding 3-nt [49][50][51]. Thus, a HJ-Lag DNA substrate with a nascent lagging-strand 30-nt longer than the nascent leading-strand was constructed ( Figure S1).
RadA/Sms·ATP or RadA/Sms C13A·ATP (10 nM) bound and unwound the nascent lagging-strand of the HJ-Lag DNA substrate with similar efficiency (Figure 2A,B, lane 4). Since forked DNA products that would be generated by coupling DNA unwinding to duplex rewinding were not observed, we discarded a branch migration activity to restore a replication fork, as observed with the RecG or RuvAB branch migrating translocases [26,33,48,52]. Instead, RadA/Sms or RadA/Sms C13A, in a first step, unwound the nascent lagging-strand. When the protein concentration was increased (e.g., to 20 nM), the 3-way junction intermediate was further processed to yield a 5 -tailed flapped intermediate and, finally, the labelled strand product (Figure 2A,B, lane 5, and S6A). It is likely that: (i) the 5 -tail of a HJ-Lag DNA penetrates the central hole of a RadA/Sms or RadA/Sms C13A hexamer; and (ii) RadA/Sms unwinds the nascent lagging-strand to yield, upon spontaneous branch migration, a suitable substrate for replication restart: a 3 -fork DNA (Figure 2A (10,15,20,30, and 40 min at 30 °C) (D). When indicated, the DNA was pre-incubated with the indicated protein (s) (5 min at 30 °C) in buffer A, then the second protein (s) and ATP were added. Reactions were stopped, deproteinized, and separated by 6% native PAGE. Gels were dried and visualized by phosphor imaging. The different orders of protein addition were separated by dashed lines. *, denotes the labelled strand; half of an arrowhead, the 3′ends; -, no protein added; B, boiled product; and I, substrate intermediates. The labelled intermediates and product were run in parallel as mobility markers. The fraction of unwound products in three independent experiments, such as those shown in (A-D), was quantified, and the mean ± SD represented.
Saturating RecA or SsbA, with respect to ssDNA, inhibited ~50% the unwinding of the nascent lagging-strand and blocked the accumulation of the radiolabeled nascent leading-strand product ( Figure 2A, lane 5 vs. lanes 12 and 16). Similar results were observed when wt RadA/Sms was replaced by RadA/Sms C13A ( Figure 2B). The inhibition is independent of a protein-protein interaction because SsbA does not interact with RadA/Sms ( Figure S2A,B) and RecA does not interact with RadA/Sms C13A [34]. It is likely that RecA or SsbA partially outcompetes RadA/Sms or RadA/Sms C13A for assembling on HJ-Lag DNA and subsequent intermediates ( Figure S6A).
To test the hypothesis, the combined action of RecA and SsbA over RadA/Sms (or RadA/Sms C13A) was analyzed by changing the order of protein addition ( Figure 2B,C). First, the HJ-Lag DNA was pre-incubated with fixed RadA/Sms (or RadA/Sms C13A) and SsbA concentrations, and then 2 mM ATP and increasing RecA concentrations were  (10,15,20,30, and 40 min at 30 • C) (D). When indicated, the DNA was pre-incubated with the indicated protein (s) (5 min at 30 • C) in buffer A, then the second protein (s) and ATP were added. Reactions were stopped, deproteinized, and separated by 6% native PAGE. Gels were dried and visualized by phosphor imaging. The different orders of protein addition were separated by dashed lines. *, denotes the labelled strand; half of an arrowhead, the 3 ends; -, no protein added; B, boiled product; and I, substrate intermediates. The labelled intermediates and product were run in parallel as mobility markers. The fraction of unwound products in three independent experiments, such as those shown in (A-D), was quantified, and the mean ± SD represented.
Saturating RecA or SsbA, with respect to ssDNA, inhibited~50% the unwinding of the nascent lagging-strand and blocked the accumulation of the radiolabeled nascent leadingstrand product (Figure 2A, lane 5 vs. lanes 12 and 16). Similar results were observed when wt RadA/Sms was replaced by RadA/Sms C13A ( Figure 2B). The inhibition is independent of a protein-protein interaction because SsbA does not interact with RadA/Sms ( Figure S2A,B) and RecA does not interact with RadA/Sms C13A [34]. It is likely that RecA or SsbA partially outcompetes RadA/Sms or RadA/Sms C13A for assembling on HJ-Lag DNA and subsequent intermediates ( Figure S6A).
To further examine the limitation on RadA/Sms unwinding of HJ-Lag DNA by SsbA or RecA, a kinetic analysis of DNA unwinding, varying the order of protein addition, at a RadA/Sms:RecA:SsbA 1:10:15 ratio, was undertaken ( Figure 2D). In a 40 min reaction, the reversed fork was fully unwound by RadA/Sms, mainly rendering the labelled nascent leading-strand, but RecA addition reduced the unwinding reaction by~3.5-fold (p < 0.01) ( Figure 2D, lane 3 vs. 4). When the DNA substrate was pre-incubated with RadA/Sms and SsbA, and then RecA and ATP were added, at min 10, only the nascent lagging-strand was unzipped ( Figure 2D, lane 5). Then, the template lagging-strand (min 15), and finally, the flayed intermediate (min 20) were unwound to render the labelled nascent leading-strand product ( Figure 2D, lanes 6-9 and Figure S5A). When RadA/Sms was pre-incubated with the substrate, and then fixed RecA, SsbA, and ATP were added, the flayed intermediate and the final product were observed at earlier times ( Figure 2D, lanes [10][11][12][13][14]. From these data, it is likely that: (i) SsbA and RecA polymerizing toward the junction reduce RadA/Sms loading on the 5 -tail of HJ-Lag DNA, and additionally inhibit RadA/Sms unwinding of the nascent lagging-strand and the intermediates; and (ii) RecA and SsbA cannot outcompete a preformed RadA/Sms-HJ-Lag DNA complex.

How
RecA Activates RadA/Sms to Unwind a HJ Structure with a Nascent Leading-Strand Longer Than the Nascent Lagging-Strand When a replication fork stalls due to a lesion on the lagging-strand, it may form a HJ-like structure with a nascent leading-strand longer than the nascent lagging-strand (HJ-Lead DNA) upon reversal ( Figure S1). It has been shown that a HJ-Lead structure, which lacks an available 5 -tail, cannot be unwound by RadA/Sms. However, RecA activates RadA/Sms to unwind its nascent lagging-strand by a poorly understood mechanism [45].
Eight different scenarios (a to g) can be envisioned to explain how RadA/Sms in concert with RecA unwinds the nascent lagging-strand of HJ-Lead DNA (SI Appendix S5): (a) traces of a contaminating 3 →5 ssDNA exonuclease in the RecA preparation may degrade the 3 tail of the HJ-Lead DNA to expose the 5 -tailed strand, with subsequent RadA/Sms self-loading; (b) dynamic RecA may partially open the duplex DNA at the junction to generate a RadA/Sms entry site (SI Appendix S6); (c) RadA/Sms may activate RecA to remodel the HJ substrate leading to flapped intermediates; (d) RecA bound to the 3 -tailed nascent leading-strand of HJ-Lead DNA might fray the nascent lagging-strand to promote RadA/Sms self-loading to the end of the nascent lagging-strand (SI Appendix S7); (e) in vivo RadA/Sms may be in a monomer-hexamer equilibrium in solution, and RecA may interact with and construct a ring-shaped RadA/Sms hexamer around the nascent lagging-strand from monomeric subunits (ring-maker strategy); (f ) in the absence of ssDNA, RadA/Sms may adopt an open-spiral hexameric structure, as reported for LonA Eco in the absence of a substrate [37], and RecA nucleated onto the ssDNA might interact with, load, and close the RadA/Sms open hexameric ring around the nascent lagging-strand (ringclosure strategy); or (g) a hexameric RecA right-handed ring [53] with a 1:1 stoichiometric relative to RadA/Sms may physically open a preformed RadA/Sms hexameric ring and load it onto the nascent lagging-strand (ring-breaker strategy), as reported for the replicative DnaB Eco helicase or the β sliding clamp [54,55]. As stated in SI Appendixes S5-S7, we have considered unlikely conditions (a) to (f ) and favored the ring-breaker hypothesis (condition g), unveiling a novel role of RecA. The sequence of the events to activate unwinding of the nascent lagging-strand-RecA complexed with hexameric RadA/Sms before interacting with the SsbA-ssDNA complex or RecA bound to ssDNA recruiting RadA/Sms on the nascent lagging-strand-is still unclear.

Two
Step RadA/Sms Loading on the Nascent-Lagging Strand of HJ-Lead DNA It has been shown that RecA·ATP neither nucleates nor polymerizes on the SsbA-ssDNA complexes [34,42]. Then, SsbA bound to the 3 -tail of the HJ-Lead DNA should indirectly inhibit RecA activation of RadA/Sms to unwind the nascent-lagging strand. Nevertheless, in the presence of RadA/Sms or RadA/Sms C13A, SsbA can be partially displaced and the ATPase activity of RecA partially recovered (Figure 1). To test whether RadA/Sms can antagonize the negative effect of SsbA on RecA binding to the 3 -tail of the HJ-Lead DNA and unwind it, a synthetic substrate with a nascent leading-strand 30-nt longer than the nascent lagging-strand was constructed (Figures 3 and S1).

Two
Step RadA/Sms Loading on the Nascent-Lagging Strand of HJ-Lead DNA It has been shown that RecA·ATP neither nucleates nor polymerizes on the SsbA-ssDNA complexes [34,42]. Then, SsbA bound to the 3′-tail of the HJ-Lead DNA should indirectly inhibit RecA activation of RadA/Sms to unwind the nascent-lagging strand. Nevertheless, in the presence of RadA/Sms or RadA/Sms C13A, SsbA can be partially displaced and the ATPase activity of RecA partially recovered (Figure 1). To test whether RadA/Sms can antagonize the negative effect of SsbA on RecA binding to the 3′-tail of the HJ-Lead DNA and unwind it, a synthetic substrate with a nascent leading-strand 30-nt longer than the nascent lagging-strand was constructed (Figures 3 and S1).  To analyze whether RadA/Sms antagonizes SsbA and facilitates RecA loading on the 3 -tailed HJ-Lead DNA, the DNA substrate was pre-incubated with fixed RadA/Sms and SsbA concentrations, and then ATP and increasing RecA were added. At a RadA/ Sms:RecA:SsbA 1:5:15 molar ratio, RadA/Sms moderately catalyzed unwinding of the nascent lagging-strand ( Figure 3A, lane 15). However, at a RadA/Sms:RecA:SsbA 1:10:15 molar ratio, the RadA/Sms unwinding efficiency was comparable to the assay in the absence of SsbA ( Figure 3A, lanes 16-17). This suggests that RadA/Sms may work as a specialized mediator provided that > 5 RecA monomers/RadA/Sms hexamer are present in the reaction.
To follow how DNA unwinding occurs, a kinetic analysis was performed with fixed RadA/Sms, RecA, and SsbA concentrations at a 1:10:15 RadA/Sms:RecA:SsbA molar ratio ( Figure 3B). RadA/Sms, in concert with RecA, unwound~95% of the HJ-Lead DNA substrate in a 40 min reaction ( Figure 3B, lane 3). When the substrate was pre-incubated with SsbA, and then RadA/Sms, RecA, and ATP were added, at 10 min, only the nascent lagging-strand was unwound, leading to a 3-way junction ( Figure 3B, lane 4). With a longer incubation time, the 5 -tailed flapped intermediate also accumulated (min 15), and at later times (min 40), the labelled nascent leading-strand was observed ( Figure 3B, lanes 5-8, Figure S6B). When the DNA substrate was pre-incubated with RadA/Sms and RecA, and then SsbA and ATP were added, at earlier times (min 10), the nascent and the template lagging-strands were unwound, and at later times (min 20), the labelled leading-strand was also unwound ( Figure 3B, lanes 9-13). In summary, the final product was formed with higher efficiency and earlier under these conditions than when SsbA was pre-incubated with the DNA (Figure 3B, lanes 9-13 vs. 4-8).
The requirement of a RadA/Sms-RecA interaction was confirmed by replacing wt RadA/Sms by RadA/Sms C13A, a variant that cannot interact with RecA [34]. Independently of the order of protein addition, the RadA/Sms C13A variant did not unwind the HJ-Lead DNA substrate ( Figure 3C, lanes 6-17).
Altogether, these data suggest that RadA/Sms-mediated unwinding of a HJ-Lead DNA is a two-step process. First, RadA/Sms counteracts the negative effect of SsbA on RecA loading onto ssDNA and facilitates RecA nucleation and filament growth on the ssDNA away from the junction. However, upon RadA/Sms-RecA interaction, RecA dynamics is inhibited, as judged by RecA-mediated ATP hydrolysis (Figure 1). Second, interaction of paused RecA·ATP or RecA·ATPγS filamented on the nascent leading-strand of HJ-Lead DNA is necessary and sufficient to load RadA/Sms onto the nascent laggingstrand, at above a RadA/Sms:RecA 1:5 stoichiometry, and unwind it from the HJ-Lead DNA substrate ( Figures 3A, S6B and S7C). To confirm that the formation of a RecA filament onto the ssDNA (and not RecA in solution) is necessary for RadA/Sms activation, an artificial substrate (a replicating fork with fully synthesized leading-and lagging-strands (3 -5 -fork DNA)) was constructed ( Figure S1). As revealed in SI Appendix S7 and Figure S7D, RecA cannot activate RadA/Sms to remodel this DNA substrate, suggesting that RecA bound to ssDNA is necessary for RadA/Sms unwinding of the nascent lagging-strand.

RecA Mediators Increase RecA Dynamics and Limit RadA/Sms Unwinding of Reversed Forks
RecO interacts with and partially displaces SsbA from ssDNA, and both proteins stimulate RecA redistribution on the ssDNA ( Figure 1C,D, line 3) [29]. In previous sections, it was shown that: (i) RadA/Sms interacts with and inhibits RecA dynamics ( Figure 1); (ii) RecA and SsbA outcompete RadA/Sms (or RadA/Sms C13A) for binding onto HJ-Lag DNA, and thereby reduce unwinding of the nascent lagging-strand ( Figure 2); (iii) RecA, filamented on the 3 tail of a HJ-Lead DNA substrate, interacts with, loads, and activates RadA/Sms to unwind the nascent lagging-strand; and (iv) depending on the protein stoichiometry and the order of protein addition, SsbA may inhibit the unwinding reaction ( Figure 3). To analyze whether the two-component mediator (RecO and SsbA), by increasing RecA dynamics, affects RadA/Sms binding and unzipping of HJ-Lag DNA, kinetic analyses of RadA/Sms-mediated unwinding, varying the order of protein addition, were undertaken ( Figure 4).
In a 40 min reaction, RadA/Sms bound to the 5 -tail of HJ-Lag DNA fully unwound the DNA substrate, but SsbA (Figure 2A, lanes 14-16), RecO, or RecA reduced the unzipping reaction ( Figure 4A, lane 4 vs. 5-6). When SsbA, RecO, and RadA/Sms were pre-incubated with HJ-Lag DNA, and then RecA and ATP were added, RadA/Sms did not unwind the substrate during the first 15 min of incubation ( Figure 4A, lanes 7-8). After 20 min of incubation, only a 3-way junction intermediate was observed (Figure 4A, lane 9). It is likely that the unwinding reaction was reduced and delayed when compared with the condition in which RecO was absent ( Figure 2D, lanes 5-7). Conversely, if RadA/Sms was pre-incubated with HJ-Lag DNA, and then SsbA, RecO, RecA, and ATP were added, the reaction occurred faster, but not to the level reached when RecO was omitted ( Figure 4A, lanes 12-16 vs. Figure 2D, lanes 10-14). It is likely that dynamic RecA, by the action of RecO-SsbA, outcompetes RadA/Sms recruitment on 5 -tailed HJ-Lag DNA more efficiently than RecA alone.  (10,15,20,30, and 40 min at 30 °C). The HJ DNA substrate was pre-incubated with the proteins denoted by 1st (5 min at 30 °C) in buffer A, then the other protein(s) and ATP were added. Reactions were stopped, deproteinized, and separated by 6% native PAGE. Gels were dried and visualized by phosphor imaging. The different orders of protein addition were separated by dashed lines. *, denotes the labelled strand; half of an arrowhead, the 3′ends; -, no protein added; B, boiled product; and I, substrate intermediates. The labelled intermediates and product were run in parallel as mobility markers. The fraction of unwound products in three independent experiments such as those shown in (A,B) was quantified, and the mean ± SD represented.
In a 40 min reaction, RadA/Sms bound to the 5′-tail of HJ-Lag DNA fully unwound the DNA substrate, but SsbA (Figure 2A, lanes 14-16), RecO, or RecA reduced the unzipping reaction ( Figure 4A, lane 4 vs. 5-6). When SsbA, RecO, and RadA/Sms were preincubated with HJ-Lag DNA, and then RecA and ATP were added, RadA/Sms did not unwind the substrate during the first 15 min of incubation ( Figure 4A, lanes 7-8). After 20 min of incubation, only a 3-way junction intermediate was observed (Figure 4A, lane 9). It is likely that the unwinding reaction was reduced and delayed when compared with the condition in which RecO was absent ( Figure 2D, lanes 5-7). Conversely, if RadA/Sms was pre-incubated with HJ-Lag DNA, and then SsbA, RecO, RecA, and ATP were added, the reaction occurred faster, but not to the level reached when RecO was omitted ( Figure  4A, lanes 12-16 vs. Figure 2D, lanes 10-14). It is likely that dynamic RecA, by the action of RecO-SsbA, outcompetes RadA/Sms recruitment on 5′-tailed HJ-Lag DNA more efficiently than RecA alone.
Then, we analyzed whether RecO and SsbA affect RecA activation of RadA/Sms-mediated unwinding of the HJ-Lead DNA ( Figure 4B). RecO did not activate RadA/Sms to unwind this substrate ( Figure 4B, lane 4). When the substrate was pre-incubated with SsbA and RecO, and then RadA/Sms, RecA, and ATP were added, the unwinding of the  (10,15,20,30, and 40 min at 30 • C). The HJ DNA substrate was pre-incubated with the proteins denoted by 1st (5 min at 30 • C) in buffer A, then the other protein(s) and ATP were added. Reactions were stopped, deproteinized, and separated by 6% native PAGE. Gels were dried and visualized by phosphor imaging. The different orders of protein addition were separated by dashed lines. *, denotes the labelled strand; half of an arrowhead, the 3 ends; -, no protein added; B, boiled product; and I, substrate intermediates. The labelled intermediates and product were run in parallel as mobility markers. The fraction of unwound products in three independent experiments such as those shown in (A,B) was quantified, and the mean ± SD represented.
Then, we analyzed whether RecO and SsbA affect RecA activation of RadA/Smsmediated unwinding of the HJ-Lead DNA ( Figure 4B). RecO did not activate RadA/Sms to unwind this substrate ( Figure 4B, lane 4). When the substrate was pre-incubated with SsbA and RecO, and then RadA/Sms, RecA, and ATP were added, the unwinding of the nascent and of the template lagging-strand was comparable to that when RecO was omitted ( Figure 4B, lanes 6-10 vs. Figure 3B, lanes 4-8). This confirms that RadA/Sms counteracts RecO-facilitated RecA dynamics (see Figure 1C,D). The preformed RadA/Sms-HJ-Lead DNA-RecA complexes unwound the nascent lagging-strand more efficiently than when SsbA and RecO were pre-bound to the DNA substrate ( Figure 4B, lanes 6-10 vs. 11-15). It is likely that: (i) RecO and SsbA limit RadA/Sms's unwinding of the nascent lagging-strand of the HJ-Lead DNA substrate, and (ii) RadA/Sms's interaction with RecA polymerizing away from the junction overrides RecA dynamics enhanced by RecO-SsbA.

RecA Activates RadA/Sms Unwinding of the Nascent Lagging-Strand of Synthetic Stalled Forks
In previous sections, it was shown that: (i) RadA/Sms binds the 5 -tail of HJ-Lag DNA and unwinds the substrate, with RecA and its accessory proteins outcompeting RadA/Sms for binding to the DNA substrate (Figures 2, 4A and S6A); and (ii) RadA/Sms, in concert with RecA, at a 1:10 molar ratio, can overcome the negative effect of SsbA, or of SsbA and RecO, on the unwinding of the nascent lagging-strand of HJ-Lead DNA substrate (Figures 1, 3, and S6B). DNA structures mimicking stalled replication forks caused by blockage on the leading-or lagging-strand ( Figure S1) were designed to: (i) gain insight into how RadA/Sms processes branched intermediates; (ii) unravel the contribution of RadA/Sms as a mediator, pausing RecA; and (iii) understand the contribution of RecA as a specialized loader.
RecA nucleated on a 15-nt long ssDNA does not hydrolyze ATP, enters into a "paused" state [50,56], and RadA/Sms poorly interacts with a 15-nt long ssDNA ( Figure S5). Then, to mimic a paused RecA and to reduce the probability of RadA/Sms self-recruitment, the length of the gapped ssDNA in the stalled fork substrate was decreased to 15-nt, either in the parental lagging-(forked-Lag) or leading-strand (forked-Lead) ( Figure S1). To confirm the integrity of the substrates, they were incubated with the branch migrating translocase RecG, which remodels a stalled fork with a leading-or a lagging-strand gap to yield a reversed fork [48]. In our assay, the nascent strands are not complementary, hence, the product should be an unreplicated fork [33]. As expected, the RecG (15 nM) control (C) unwound~50% of nascent strands to render an unreplicated fork product with similar efficiency in the two substrates ( Figure 5A,B, lane 18).
RecA, SsbA, or RadA/Sms neither reversed the forked-Lag or forked-Lead DNA substrate nor unwound their nascent strands ( Figure 5A,B, lanes 3-5). In the presence of RecA, however, RadA/Sms unwound the nascent lagging-strand of both substrates with similar efficiency (Figure 5A,B, lanes 6-8). When wt RadA/Sms is replaced by RadA/Sms C13A, no unwinding was observed. It is likely that: (i) RecA nucleated toward the junction on the gapped forked-Lag DNA or away from the junction on the gapped forked-Lead DNA is necessary and sufficient to activate RadA/Sms to unwind the nascent lagging-strand of both DNA substrates, and (ii) RecA interacts with and modulates the conformation of hexameric RadA/Sms and loads it in cis or in trans on the nascent lagging-strand ( Figure S6C,D).
To analyze whether SsbA binds to and inhibits RecA nucleation onto the ssDNA gap and indirectly blocks RecA activation of RadA/Sms-mediated unwinding of these stalled fork substrates, and if RadA/Sms alone or in concert with RecA antagonizes the negative effect of SsbA on RecA nucleation, the order of protein addition was varied. When a preformed SsbA-forked-Lag or SsbA-forked-Lead complex was incubated with RadA/Sms and RecA (at a RadA/Sms:RecA:SsbA 1:5:15 stoichiometry), RadA/Sms-mediated unzipping of the nascent lagging-strand was not observed ( Figure 5A,B, lane 9). Increasing RecA concentrations, to RadA/Sms:RecA:SsbA 1:10:15 and 1:20:15 molar ratios, were necessary to counteract the negative effect of SsbA on RadA/Sms unwinding of the nascent laggingstrand of both DNA substrates ( Figure 5A,B, lanes 10-11). It is likely that a RecA-RadA/Sms stoichiometry of 1:10 or 1:20 is necessary and sufficient to antagonize the negative effect of SsbA on RecA nucleation and subsequent loading of RadA/Sms. In some reactions, the forked DNA was pre-incubated with the indicated protein (s) (5 min at 30 °C) in buffer A, then the second protein (s) and ATP were added. Reactions were stopped, deproteinized, and separated by 6% native PAGE. Gels were dried and visualized by phosphor imaging. The different orders of protein addition were separated by dashed lines. *, denotes the labelled strand; half of an arrowhead, the 3′ends; -, no protein added; B, boiled product; C, a control reaction with RecG (15 nM), which branch migrates the DNA substrate, resulting in unreplicated forked DNA; and I, substrate intermediates. The labelled intermediates and product were run in parallel as mobility markers. The fraction of unwound products in three independent experiments such as those shown in (A,B) was quantified, and the mean ± SD represented.
RecA, SsbA, or RadA/Sms neither reversed the forked-Lag or forked-Lead DNA substrate nor unwound their nascent strands ( Figure 5A,B, lanes 3-5). In the presence of RecA, however, RadA/Sms unwound the nascent lagging-strand of both substrates with similar efficiency ( Figure 5A,B, lanes 6-8). When wt RadA/Sms is replaced by RadA/Sms C13A, no unwinding was observed. It is likely that: (i) RecA nucleated toward the junction on the gapped forked-Lag DNA or away from the junction on the gapped forked-Lead DNA is necessary and sufficient to activate RadA/Sms to unwind the nascent lagging-strand of both DNA substrates, and (ii) RecA interacts with and modulates the conformation of hexameric RadA/Sms and loads it in cis or in trans on the nascent lagging-strand ( Figure  S6C,D).
To analyze whether SsbA binds to and inhibits RecA nucleation onto the ssDNA gap and indirectly blocks RecA activation of RadA/Sms-mediated unwinding of these stalled fork substrates, and if RadA/Sms alone or in concert with RecA antagonizes the negative In some reactions, the forked DNA was pre-incubated with the indicated protein (s) (5 min at 30 • C) in buffer A, then the second protein (s) and ATP were added. Reactions were stopped, deproteinized, and separated by 6% native PAGE. Gels were dried and visualized by phosphor imaging. The different orders of protein addition were separated by dashed lines. *, denotes the labelled strand; half of an arrowhead, the 3 ends; -, no protein added; B, boiled product; C, a control reaction with RecG (15 nM), which branch migrates the DNA substrate, resulting in unreplicated forked DNA; and I, substrate intermediates. The labelled intermediates and product were run in parallel as mobility markers. The fraction of unwound products in three independent experiments such as those shown in (A,B) was quantified, and the mean ± SD represented.
When SsbA was added to preformed RecA-forked-Lag-RadA/Sms or RecA-forked-Lead-RadA/Sms complexes, the rate of RadA/Sms-mediated unwinding was similar to that in the absence of SsbA ( Figure 5A,B, lanes 6-8 and 12-14). This suggests that SsbA cannot act against RadA/Sms-RecA bound to forked-Lag or forked-Lead DNA. Finally, when SsbA, RadA/Sms, and forked-Lag or forked-Lead DNA were pre-incubated, and then fixed ATP and increasing RecA concentrations were added, RadA/Sms-mediated unwinding of the nascent lagging-strand was not significantly different when compared with the condition of RadA/Sms and RecA first ( Figure 5A,B, lanes 15-17 vs. [12][13][14]. It is likely that RadA/Sms facilitates the displacement of SsbA, with RecA being necessary to load RadA/Sms on the nascent lagging-strand. From the data presented in Figure 5, we can predict ( Figure S6C,D) that: (i) RecA nucleates on a 15-nt gap, but it cannot remodel the small synthetic stalled forks; (ii) RadA/Sms neither remodels nor unwinds the synthetic forked-Lag or forked-Lead DNA substrate; (iii) the 5 -end of nucleated RecA on a 15-nt long ssDNA gap is necessary and sufficient to direct RadA/Sms, both in cis or in trans, to their nascent lagging-strand; (iv) saturating SsbA bound to gapped ssDNA, perhaps in the SsbA 15

Discussion
B. subtilis respond to a replicative stress by forming RecA foci, in concert with SsbA and RecO, that colocalize with the replisome (colocalization > 90%) at a stalled replication fork [19,22]. RadA/Sms forms dynamic foci in wt cells but static foci in ∆recG cells, a mutation that causes accumulation of branched intermediates [38,57]. In this study, we sought to decipher whether paused RecA and ring-shaped homohexameric RadA/Sms define a novel function upon replication stress to promote replication restart. In summary, first, RecA interacts with and loads RadA/Sms at stalled forks [34,45]. Second, RadA/Sms inhibits RecA-mediated ATP hydrolysis, crucial for its dynamic activity ( Figure 1). Third, RadA/Sms loaded at the nascent lagging-strand of stalled or reversed forks unwinds it to generate, directly or indirectly, a 3 -fork DNA substrate (Figures 3 and 5). Fourth, the pre-primosome PriA-DnaD-DnaB proteins recognize the 3 -fork DNA, and with DnaI, load the replicative DnaC helicase for replication restart [46]. In fact, in the absence of RecA, recruitment of pre-preprimosome DnaD and replicative DNA helicase DnaC is significantly reduced [23].
From previous studies and the data presented here, six ideas can be drawn. First, cytological analyses have shown that SsbA, when travelling with the replisome, may load RecO, RecG, or PriA at a stalled replication fork [17,18]. RecO contributes to RecA foci formation at stalled replication forks [19,22], RecG is a genuine fork remodeler, and PriA is required at a later stage for replication re-initiation [46,48].
Second, SsbA interacts with and loads RecO onto a lesion-containing gap, then RecO partially displaces SsbA and generates the substrate for RecA nucleation [29]. Here, RecA, in concert with its mediators and modulators, would catalyze DNA invasion and strand transfer, crucial for template switching and double-strand break repair [58]. Alternatively, RadA/Sms interacts with and limits RecA-mediated ATP hydrolysis ( Figure 1) and strand transfer ( Figure S4). This way, RadA/Sms may prevent RecA from provoking unnecessary DNA strand exchange at stalled or reversed forks.
Third, RecA neither nucleates nor polymerizes on the SsbA-ssDNA complexes [28][29][30]. However, RadA/Sms, as a specialized mediator, stimulates RecA nucleation onto ssDNA. Concretely, RadA/Sms counteracts the negative effect of SsbA on RecA nucleation on the 30-nt long 3 -tail of a HJ-Lead DNA substrate, or on the 15-nt gap at the leading or lagging template strand of a forked-Lead or forked-Lag DNA, and limits RecA redistribution, even in the presence of the positive RecO mediator (Figures 1, 3, and 6A,B). It is unclear whether RadA/Sms alone or upon a transient interaction with RecA undergoes a structural change to displace SsbA and load RecA. At present, we cannot rule out that RadA/Sms modulates RecA conformation to a form that can displace SsbA from ssDNA, although we favor the former hypothesis.
Fourth, RecA·ATP (or RecA·ATPγS) protects the nascent strands from degradation and loads RadA/Sms on the nascent lagging-strand of stalled forks (or of reversed forks with a nascent leading-strand longer than the nascent lagging-strand) by a poorly understood mechanism ( Figure 6A,B,D). RecA·ATP forms right-handed helical filaments of 6.2 RecA protomers per helix turn on ssDNA, extending it by 50% relative to B-form dsDNA [51,53]. If each RecA monomer interacts with each protomer of RadA/Sms, the hexameric form of the latter may open and ssDNA may pass through the gap in the ring. Then, RecA would deliver RadA/Sms in cis or in trans onto the nascent lagging-strand to render a 3 -fork DNA specific for replication restart. This mechanism mimics loading of the replicative helicase DnaB Eco by DnaC Eco self-oligomerized as a right-handed helical filament by a ring-breaking strategy [54]. structural change to displace SsbA and load RecA. At present, we cannot rule out that RadA/Sms modulates RecA conformation to a form that can displace SsbA from ssDNA, although we favor the former hypothesis.
Fourth, RecA·ATP (or RecA·ATPγS) protects the nascent strands from degradation and loads RadA/Sms on the nascent lagging-strand of stalled forks (or of reversed forks with a nascent leading-strand longer than the nascent lagging-strand) by a poorly understood mechanism ( Figures 6A,B,D). RecA·ATP forms right-handed helical filaments of 6.2 RecA protomers per helix turn on ssDNA, extending it by 50% relative to B-form dsDNA [51,53]. If each RecA monomer interacts with each protomer of RadA/Sms, the hexameric form of the latter may open and ssDNA may pass through the gap in the ring. Then, RecA would deliver RadA/Sms in cis or in trans onto the nascent lagging-strand to render a 3′fork DNA specific for replication restart. This mechanism mimics loading of the replicative helicase DnaBEco by DnaCEco self-oligomerized as a right-handed helical filament by a ring-breaking strategy [54].
Fifth, activated RadA/Sms unwinds the nascent lagging-strand of stalled forks (Figure 6A,B) or of reversed forks (HJ-Lead DNA) ( Figure 6D) to generate, directly or indirectly, a 3′-fork DNA intermediate suitable to be recognized by the pre-primosomal proteins, an essential step for replication re-start [46]. Here, RecA and its accessory proteins limit further RadA/Sms-mediated unwinding of the 3′-fork DNA.  Fifth, activated RadA/Sms unwinds the nascent lagging-strand of stalled forks ( Figure 6A,B) or of reversed forks (HJ-Lead DNA) ( Figure 6D) to generate, directly or indirectly, a 3 -fork DNA intermediate suitable to be recognized by the pre-primosomal proteins, an essential step for replication re-start [46]. Here, RecA and its accessory proteins limit further RadA/Sms-mediated unwinding of the 3 -fork DNA.
Finally, RadA/Sms bound to the 5 tail of HJ-Lag DNA unwinds the nascent laggingstrand to render a 3 -fork DNA substrate upon spontaneous remodeling (Figures 2 and S6A,C). Here, RecA and the SsbA and RecO mediators limit RadA/Sms binding onto and unwinding of the nascent lagging-strand to prevent incorrect unwinding of the template lagging-strand (Figures 2 and S6A,C).
In conclusion, as reported in other systems, in B. subtilis, different motor proteins process replicative stress intermediates [48]. However, unlike the fork remodeler RecG or RuvAB, RadA/Sms seems to selectively unwind the nascent lagging-strand of stalled or reversed forks. This way, a branched DNA intermediate that resembles a proper substrate for PriA-dependent replication restart is generated.

ATP Hydrolysis Assays
The ATP hydrolysis activity of RecA, RadA/Sms (or of its mutant variants), and PcrA was assayed in the presence of RecA mediators via an NAD/NADH coupled spectrophotometric enzymatic assay [47]. Since RadA/Sms and its variants were purified using the same protocol, and the RadA/Sms K104R and RadA/Sms K255R variants show no ATP hydrolysis, we are confident that the measured ATPase activity is associated to the RadA/Sms or RadA/Sms C13A protein.
The rate of ATP hydrolysis was measured in buffer A containing 5 mM ATP and an ATP regeneration system (620 µM NADH, 100 U/mL lactic dehydrogenase, 500 U/mL pyruvate kinase, and 2.5 mM phosphoenolpyruvate) for 30 min at 37 • C [47]. The order of addition of circular 3199-nt pGEM3 Zf(+) (cssDNA) (10 µM in nt) and of the purified proteins (and their concentration) is indicated in the text. Data obtained from A 340 absorbance were converted to ADP concentrations and plotted as a function of time [47]. Accepting that RecA, wt RadA/Sms, or its mutant variant RadA/Sms C13A operates mostly under steadystate conditions, the maximal number of substrate-to-product conversion per unit of time (V max ) was measured. t-tests were applied to analyze the statistical significance of the data. At longer times, ATP hydrolysis seems to decrease, but this is solved by increasing the concentration of the ATP regeneration system, suggesting that at very late times it is rate limiting.

DNA Unwinding Assays
The different forked DNA substrates (0.5 nM) used were incubated with increasing concentrations of RadA/Sms, its mutant variant RadA C13A, or RecG in the presence of the indicated concentrations of RecA, RecO, or SsbA for 15 min at 30 • C in buffer A containing 2 mM ATP in a 20-µL volume. The reactions were deproteinized by phenolchloroform; DNA substrates and products were precipitated by NaCl and ethanol addition and subsequently separated using 6% (w/v) PAGE. Gels were run and dried prior to phosphor-imaging analysis. The bands were quantified using ImageJ (NIH). t-tests were applied to analyze the statistical significance of the data.

RecA-Mediated DNA Strand Exchange
The 3199 bp KpnI-cleaved pGEM3 Zf(+) dsDNA (20 µM in nt) and the homologous circular 3199-nt ssDNA (10 µM in nt) were pre-incubated with the indicated protein or protein combination in buffer A containing 5 mM ATP or dATP (5 min at 37 • C). Then, a fixed RecA and a fixed or variable wt or mutant RadA/Sms concentration were added and the reaction was incubated for fixed or variable times at 37 • C. An ATP regeneration system (8 units/mL creatine phosphokinase and 8 mM phosphocreatine) was included in the recombination reaction. After incubation of the recombination reaction, samples were de-proteinized as described [65] and fractionated through 0.8% agarose gel electrophoresis with ethidium bromide. The signal of the remaining linear dsDNA (lds) substrate and the appearance of joint molecule (jm) intermediates and nicked circular (nc) products were quantified from the gels using a Geldoc system and the Image Lab software (BioRad, Hercules, CA, USA), as described [30]. When indicated, the sum of jm and nc are shown (% of recombination).

Protein-Protein Interaction Assays
In vitro protein-protein interactions were assayed using affinity chromatography as earlier described [45]. His-tagged RadA/Sms, native SsbA, or RadA/Sms and SsbA (1 µg each) was (were) pre-incubated in buffer A (5 min, 37 • C), both in the presence or absence of 10 µM cssDNA and 5 mM ATP, and then loaded onto a 50 µL Ni 2+ matrix microcolumn, and the FT was collected. After extensive washing, the retained proteins in the Ni 2+ matrix were eluted with 50 µL of buffer A containing 400 mM imidazole. The proteins were separated by 10% SDS-PAGE, and gels were stained with Coomassie Blue (Merck, Darmstadt, Germany).