The Escherichia coli DinD Protein Modulates RecA Activity by Inhibiting Postsynaptic RecA Filaments*

Escherichia coli dinD is an SOS gene up-regulated in response to DNA damage. We find that the purified DinD protein is a novel inhibitor of RecA-mediated DNA strand exchange activities. Most modulators of RecA protein activity act by controlling the amount of RecA protein bound to single-stranded DNA by affecting either the loading of RecA protein onto DNA or the disassembly of RecA nucleoprotein filaments bound to single-stranded DNA. The DinD protein, however, acts postsynaptically to inhibit RecA during an on-going DNA strand exchange, likely through the disassembly of RecA filaments. DinD protein does not affect RecA single-stranded DNA filaments but efficiently disassembles RecA when bound to two or more DNA strands, effectively halting RecA-mediated branch migration. By utilizing a nonspecific duplex DNA-binding protein, YebG, we show that the DinD effect is not simply due to duplex DNA sequestration. We present a model suggesting that the negative effects of DinD protein are targeted to a specific conformational state of the RecA protein and discuss the potential role of DinD protein in the regulation of recombinational DNA repair.

The Escherichia coli SOS response is a coordinately regulated network of genes induced in reaction to heavy or persistent DNA damage (1). The SOS regulon is repressed by the LexA protein, which is inactivated as a repressor upon cellular stress such as heavy DNA damage. The cellular signal for SOS induction is the RecA protein bound to single-stranded DNA (ssDNA). 2 RecA is the central DNA recombinase in bacteria and carries out several distinct functions when activated (2). The most appreciated function of RecA is the catalysis of homologous DNA recombination crucial to the generation of genetic diversity. However, based on frequency of use, the primary role of RecA lies in the multiple pathways for the recombinational repair of stalled DNA replication forks. Through much in vitro work, it has become increasingly apparent that the RecA protein is under considerable control by a network of proteins that function to modulate when and where RecA protein binds to DNA (3). The PsiB protein has recently been shown to bind to free RecA protein, effectively inhibiting RecA from nucleating onto ssDNA (4). RecA is also inhibited from nucleating onto ssDNA bound by the single-stranded DNA-binding protein (SSB) (5). The SSB-imposed inhibition is relieved by the action of the RecF, RecO, and RecR proteins (3). Following nucleation, RecA protein protomers assemble into a nucleoprotein filament, extending cooperatively in the 5Ј to 3Ј direction. The RdgC protein can inhibit RecA binding to ssDNA and can interfere with homologous DNA pairing by binding to duplex DNA (6,7). Some proteins shown to dismantle RecA filaments are known DNA translocases, such as UvrD (8) and PcrA helicases (9). Filament extension can be blocked through the action of the RecX protein (10), whereas the DinI protein antagonizes the function of RecX by stabilizing RecA filaments, inhibiting filament end-dependent disassembly (11,12). The phenotypes of both recX and dinI mutants are subtle. But, in agreement with the biochemical data, the number of RecA-GFP foci observed in live cells is higher in recX mutants and lower in both the dinI single and dinI, recX double mutants (13). The expression of both the DinI and RecX proteins is controlled by the LexA repressor (14,15), defining their membership in the SOS regulon. There appears to be a network of proteins that function to regulate RecA filaments in bacterial systems. Because the mutant phenotypes of RecA regulatory genes are frequently subtle, at best, it may be difficult to employ typical genetic screens to identify other potential RecA mediator genes. The E. coli genome consists of hundreds of genes that are expressed in response to DNA damage, and approximately 40 of those genes are directly controlled by the LexA repressor (16,17). We have endeavored to ascribe molecular roles to several of these genes for which there is no described or predicted function. We started by investigating the DinD and YebG proteins.
The damage-inducible protein D (DinD) and YebG proteins are part of the SOS regulon and are highly induced upon DNA damage. The yebG gene was identified based on the similarity of its promoter sequence to known LexA boxes (18). The DNA damage-dependent induction of yebG is dependent on the cya gene (19). The dinD gene (also known as pcsA (20)) was first identified in a screen for DNA damage inducible genes (14). Later, the gene was described again based on a cold-sensitive mutant that forms filamentous cells (20). The cold-sensitive phenotype is due to a toxic dinD point mutant, pcsA68, which is suppressed by either a ⌬recA or ⌬dinD mutation. This study also suggested that the dinD gene is under the transcriptional control of the LexA repressor protein, which was independently determined using a combinatorial screen (21). In fact, the dinD promoter is frequently utilized as a fusion for reporters of SOS (14,22,23) presumably because dinD is expressed during the SOS response and its expression is increased nearly 10-fold in response to UV irradiation (16). The E. coli DinD and YebG proteins consist of 274 and 96 amino acid residues, respectively. Bioinformatics have not revealed any obvious sequence homologs of either YebG or DinD and the sequence contains no predictable motifs that might suggest function. Both proteins are substrates for the ClpP protease system, a characteristic shared by several SOS-regulated gene products (24). Although we describe basic biochemical characteristics for the YebG protein, the focus of the current study is a novel RecA regulatory mechanism that we have observed for the E. coli DinD protein.

EXPERIMENTAL PROCEDURES
Enzymes and Biochemicals-The E. coli RecA protein was purified as described (25). The concentration of purified RecA protein was determined from the absorbance at 280 nm using the extinction coefficient 2.23 ϫ 10 4 M Ϫ1 cm Ϫ1 (26). E. coli SSB was purified as described (27). The concentration of the purified SSB protein was determined from the absorbance at 280 nm using the extinction coefficient of 2.83 ϫ 10 4 M Ϫ1 cm Ϫ1 (28). Unless otherwise noted, all of the reagents were purchased from Fisher. ATP regeneration and coupling components were from Sigma.
DNA Substrates-Bacteriophage X174 circular ssDNA (virion) and X174 RF I supercoiled circular duplex DNA were purchased from New England Biolabs. Full-length linear duplex DNA was generated by digestion of X174 RF I DNA (5,386 bp) with the PstI restriction endonuclease, using conditions suggested by the enzyme supplier (New England Biolabs). The digested DNA was extracted with phenol/chloroform/isoamyl alcohol (25:24:1), followed by ethanol precipitation. Circular X174 duplex DNA containing a single nick was generated using a previously described method (29). Circular ssDNA from bacteriophage M13mp8 (7,229 nucleotides) was prepared as previously described (30). The concentrations of ssDNA and double-stranded DNA (dsDNA) were determined by absorbance at 260 nm, using 36 and 50 g ml Ϫ1 A 260 Ϫ1 , respectively, as conversion factors. Linear single-stranded poly(dT), average length of 229 nucleotides, was purchased from GE Healthcare. All DNA concentrations are given in micromolar nucleotides.
Expression and Purification of the DinD Protein-The nuclease-deficient E. coli strain STL2669 (a gift from Susan T. Lovett of Brandeis University (25)) was co-transformed with pT7pol26 (31) and pEAW439. This latter plasmid is a pET21D derivative carrying the E. coli dinD gene (from MG1655) under control of the T7 RNA polymerase promoter. A 10-liter culture was grown to an A 600 of 0.5 in LB broth (10 g/liter of tryptone, 5 g/liter of yeast extract, and 10 g/liter of NaCl, with pH adjusted to 7.0). DinD protein expression was induced by 0.4 mM (final concentration) isopropyl 1-thio-␤-D-galactopyranoside (Gold Biotechnology). Following a 3-h incubation at 37°C, cells were harvested by centrifugation, flash-frozen in liquid N 2 , and stored at Ϫ80°C.
The following DinD protein purification steps were carried out at 4°C. Cell paste (15 g) was thawed overnight to a 20% (w/v) cell ratio in Tris sucrose solution (25% (w/v) sucrose and 250 mM Tris-HCl (80% cation, pH 7.5)). Lysozyme solution (in 250 mM Tris-HCl, 80% cation) was added (2.5 mg/ml final concentration) to cells and incubated for 60 min, followed by the addition of 0.4 ml of 25 mM EDTA/ml of lysed cell suspension, sonication, and centrifugation. DNA was precipitated out of the resulting soluble fraction by incubation with polyethyleneimine (0.5% final concentration, pH 7.5) for 60 min, followed by centrifugation. The DinD protein was precipitated from the supernatant by the addition of 0.33 g of solid ammonium sulfate/ml of solution (55% saturation). The precipitate, after centrifugation, was washed two times with R buffer (20 mM Tris-HCl (80% cation, pH 7.5), 1 mM dithiothreitol (DTT), 0.1 mM EDTA, and 10% (w/v) glycerol) ϩ 0.35 g/ml of ammonium sulfate, and resuspended in R buffer ϩ 100 mM KCl. The solution was dialyzed twice versus R buffer ϩ 100 mM KCl and loaded onto a DEAE-Sepharose (GE Healthcare) column. The column was washed with R buffer ϩ 100 mM KCl and flow-through fractions were analyzed by SDS-PAGE. The fractions containing DinD protein were pooled and dialyzed versus P buffer (20 mM potassium phosphate, pH 6.8, 1 mM DTT, 0.1 mM EDTA, and 10% (w/v) glycerol) ϩ 150 mM KCl and loaded onto a ceramic hydroxyapatite column (Bio-Rad), washed with 2 column volumes of P buffer ϩ 150 mM KCl, and eluted with a linear gradient from 20 mM to 1 M potassium phosphate buffer ϩ 150 mM KCl, pH 6.8, over 10 column volumes. The fractions containing DinD protein were pooled and dialyzed extensively versus storage buffer (R buffer ϩ 100 mM KCl), determined to be free of nuclease contamination, flash-frozen in liquid N 2 , and stored at Ϫ80°C. The DinD protein concentration (molecular mass 31,078 Da) was determined from the absorbance at 280 nM using the native extinction coefficient 19,427 M Ϫ1 cm Ϫ1 determined as described (32).
Expression and Purification of the YebG Protein-E. coli strain STL2669 was co-transformed with pT7pol26 (31) and pEAW441. This latter plasmid is a pET21A derivative carrying the E. coli yebG gene (from MG1655) under control of the T7 RNA polymerase promoter. A 10-liter culture was grown to an A 600 of 0.5 in LB broth. YebG protein expression was induced as described above for the DinD protein.
The YebG protein was purified using the protocol detailed above for the DinD protein except for the following. The YebG protein was precipitated out of the lysate by polyethyleneimine (0.5% final concentration, pH 7.5) addition followed by a 60-min incubation and centrifugation. The pellet was washed twice with R buffer ϩ 150 mM ammonium sulfate and extracted two times with R buffer ϩ 300 mM ammonium sulfate. The protein solution was precipitated by the addition of 0.26 g of solid ammonium sulfate/ml of solution (45% saturation). The precipitant was washed twice with R buffer ϩ 0.28 g/ml of ammonium sulfate, resuspended in R buffer ϩ 200 mM KCl, and dialyzed once versus R buffer ϩ 200 mM KCl, then twice versus R buffer ϩ 50 mM KCl. The protein was loaded onto a DEAE-Sepharose column, washed with two column volumes of R buffer ϩ 50 mM KCl, and eluted with a linear gradient of KCl from 50 mM to 1 M KCl over 10 column volumes. Fractions containing YebG protein were pooled and dialyzed versus P buffer. The dialyzed solution was loaded onto a ceramic hydroxyapatite column, washed with 2 column volumes of P buffer, and eluted with a linear gradient from 20 mM to 1 M potassium-phosphate buffer, pH 6.8, over 10 column volumes. The fractions containing YebG protein were pooled and dialyzed extensively versus R buffer ϩ 50 mM KCl (storage buffer), determined to be free of nuclease contamination, flash-frozen in liquid N 2 , and stored at Ϫ80°C. The concentration of the YebG protein (molecular mass 10,717 Da) was determined from the absorbance at 280 nm using the calculated extinction coefficient 13,940 M Ϫ1 cm Ϫ1 .
ATP Assay-The DNA-dependent ATPase activity of the RecA protein was measured using an enzyme-coupled spectrophotometric enzyme assay (33,34) carried out with a Varian Cary 100 Bio-dual beam spectrophotometer equipped with a temperature controller and a 12-position cell changer. The cell path length was 1 cm. The regeneration of ATP from phosphoenolpyruvate and ADP was coupled to the oxidation of NADH through the addition of pyruvate kinase and lactate dehydrogenase. The decrease in absorbance of NADH at 380 nm (extinction coefficient of 1.21 mM Ϫ1 cm Ϫ1 ) was used to calculate the rate of ATP hydrolysis. The reactions were carried out at 37°C in 25 mM Tris acetate (80% cation, pH 7.4), 1 mM DTT, 3 mM potassium glutamate, 10 mM Mg(OAc) 2 , 5% (w/v) glycerol, an ATP regeneration system (10 units/ml of pyruvate kinase and 3 mM phosphoenolpyruvate), and a coupling system (1.5 mM NADH and 10 units/ml of lactate dehydrogenase). The concentration of DNA, RecA, and DinD proteins, where applicable, as well as the order of protein additions are indicated in the figure legends. Reactions were initiated by the addition of 3 mM ATP and SSB protein 10 min after the addition of RecA protein, unless otherwise noted. The concentration of SSB protein used is 0.1 ϫ ssDNA concentration in all experiments.
RecA-mediated DNA Three-strand Exchange-The DNA strand exchange reactions were carried out at 37°C in 25 mM Tris acetate (80% cation, pH 7.4), 1 mM DTT, 3 mM potassium glutamate, 10 mM Mg(OAc) 2 , 5% (w/v) glycerol, and an ATP regeneration system (10 units/ml of pyruvate kinase and 2.5 mM phosphoenolpyruvate). RecA protein (6.7 M) was incubated with 20 M X174 circular ssDNA for 10 min. SSB protein (2 M) and 3 mM ATP were added, followed by another 10-min incubation. The reaction was initiated by the addition of 20 M X174 linear duplex DNA and incubated for 60 min, or the times indicated in the figure legend. Reactions were stopped by adding a 2ϫ solution containing 15% Ficoll, 4% SDS, 0.24% bromphenol blue, and 0.24% xylene cyanole, and proteinase K (to 1.25 mg/ml). Samples were subjected to electrophoresis in 0.8% agarose gels with 1 ϫ TBE buffer (90 mM Tris borate, 2 mM EDTA, pH 8), stained with ethidium bromide, and exposed to UV light. The inverted gel images were obtained using a digital CCD camera with Foto/Analyst Pc Image software version 10.21 (Fotodyne). The DinD or YebG proteins were included to this basic reaction setup in various orders of addition, as detailed in the figure legends.
ATP hydrolysis was measured during the RecA-mediated DNA strand exchange reaction for some experiments. In this case, strand exchange reactions were set up as described above except that the coupling system of lactate dehydrogenase and NADH described for the ATPase assays above was also included. Furthermore, the concentration of X174 circular ssDNA and linear duplex DNA were reduced to 4 and 8 M nucleotides, respectively. The concentration of DNA, RecA and DinD, or YebG proteins, where applicable, as well as the order of protein additions are indicated in the figure legends. For all experiments in which the state of RecA filaments during DNA strand exchange was monitored by measuring ATP hydrolysis, corresponding reactions of the same RecA and DNA ratios were also monitored by agarose gel electrophoresis.
DNA Binding-DNA binding by the DinD or YebG proteins was assessed using an electrophoretic mobility shift assay (EMSA). The binding reaction buffer consisted of 25 mM Tris acetate (80% cation, pH 7.4), 1 mM DTT, 3 mM potassium glutamate, 10 mM Mg(OAc) 2 , 5% (w/v) glycerol, and 1 mg/ml of BSA. Protein was incubated with 20 M X174 circular ssDNA or 20 M X174 linear duplex DNA at 37°C for 30 min for YebG. The DinD protein was incubated for 5 min at 37°C followed by 15 min on ice. Ficoll (to 4%) was added and the reactions were subjected to electrophoresis in 1% agarose with 50 mM glycine buffer for DinD or 0.6% agarose with 1 ϫ TAE buffer (40 mM Tris acetate, 1 mM EDTA, pH 8) for YebG. Electrophoresis was conducted at 25°C at 100 V. After electrophoresis, the gels were stained with SYBR Gold (Invitrogen) for ssDNA or ethidium bromide for dsDNA and exposed to UV light. The inverted gel images were obtained using a digital CCD camera with Foto/Analyst Pc Image software version 10.21 (Fotodyne).

The DinD and YebG Proteins Are Nonspecific DNA-binding
Proteins-We set out in this study to define the molecular role of the SOS-regulated DinD and YebG proteins. We purified recombinant native (containing no affinity tag) E. coli DinD and YebG proteins to near homogeneity. We confirmed that the DinD protein can bind to both circular ssDNA (Fig. 1A) and linearized duplex DNA (Fig. 1B) by using an electrophoretic mobility shift assay. The reduced mobility of the DNA in the agarose gel is indicative of a DinD-DNA complex. The DNA molecules used for the binding assays shown in Fig. 1 were derived from bacteriophage X174 (see "Experimental Procedures"). However, we have also observed binding to several other duplex DNA substrates derived from plasmid DNA and poly(dT) linear ssDNA (data not shown). The latter confirms that the ssDNA binding activity shown in Fig. 1A is a not a result of DinD binding to regions of secondary structures that may exist in the single-stranded virion DNA derived from bacteriophage. We conclude that DinD is a DNA-binding protein capable of binding to both ssDNA and duplex DNA nonspecifically with respect to sequence.
Additionally, we observed that YebG interacts with duplex DNA (Fig. 1C). YebG can bind duplex DNA molecules derived from various sources, such as bacteriophage X174 (Fig. 1C) and plasmid DNA (data not shown), indicating that YebG also binds to duplex DNA nonspecifically with respect to sequence. We have not observed YebG protein binding to ssDNA under multiple conditions (data not shown).
The DinD Protein Can Interfere with RecA Nucleation onto ssDNA, but Has No Significant Effect on RecA Filaments Formed on ssDNA-The rate of ATP hydrolyzed by the RecA protein is dependent on the amount of RecA bound to ssDNA (2). To assess the potential impact of DinD on RecA filaments, we measured RecA filament assembly onto circular ssDNA preincubated with the DinD protein by monitoring the ATP hydrolytic activity of RecA protein. DinD protein does not hydrolyze ATP under multiple conditions tested (data not shown). The RecA protein (2 M) exhibits a lag (defined as the time required to reach a steady-state rate of DNA-dependent ATP hydrolysis) to binding 5 M circular ssDNA when as little as 250 nM DinD is preincubated with the DNA compared with RecA assembly in the absence of DinD ( Fig. 2A). However, the RecA-catalyzed ATP hydrolysis recovers over time (depending on the DinD concentration) indicating that RecA can overcome the DinD inhibition, possibly by displacing DinD. This suggests that the DinD and RecA proteins compete for DNA binding sites. Therefore, we added DinD protein to RecA filaments preformed on 4 M circular ssDNA and measured the ATP hydrolytic activity of those filaments. Surprisingly, the addition of DinD had no significant effect on the ssDNA-dependent ATP hydrolysis activity of 2 M RecA protein even when an excess of DinD protein (up to 4 M) is added to those filaments ( Fig. 2B). Even RecA nucleoprotein filaments formed with subsaturating RecA protein to DNA (1.5 M RecA, 5 M circular ssDNA) were not disrupted by the addition of a 3-fold excess of DinD to RecA protein (data not shown). Therefore, DinD can exert an impediment to RecA filaments only when added prior RecA filament assembly.
DinD Protein Inhibits the RecA Protein-mediated DNA Strand Exchange Reaction-Because several LexA-regulated gene products have profound effects on the function of the RecA protein, we tested whether the purified DinD protein had any effect on the in vitro RecA-mediated DNA strand exchange reaction diagrammed in Fig. 3. This reaction is thought to mimic the central steps of recombinational DNA repair of stalled replication forks carried out by RecA protein in vivo (2). The assay for DNA strand exchange is relatively modular in its design. The discrete stages of the RecA-mediated strand exchange reaction can be monitored by agarose gel electrophoresis to assess the state of the DNA substrates and products. Furthermore, the state of the RecA filaments can be evaluated by measuring ATP hydrolysis in real-time as the reaction pro-  gresses (35,36). In the first phase of the reaction, RecA protein and ATP form continuous filaments on circular ssDNA with a stoichiometry of 1 RecA monomer per 3 nucleotides (2). These presynaptic RecA filaments hydrolyze ATP throughout the length of the filament with a turnover rate of about 30 min Ϫ1 . DNA synapsis is initiated in the second phase of the DNA strand exchange reaction by the addition of homologous linear duplex DNA. The rate of ATP hydrolysis by the RecA protein abruptly drops by ϳ30% at this step (37). In the final intermediate pairing and branch migration phase of the reaction, RecA filaments are in contact with three strands of DNA. The strand exchange reaction concludes with the formation of nicked circular heteroduplex DNA molecules and the RecA filament is now associated with that product duplex (38,39). The inclusion of the DinD protein under the standard conditions described under "Experimental Procedures" resulted in the inhibition of this reaction, as detailed below (Fig. 4). We have taken advantage of the modular nature of the DNA strand exchange reaction to identify a novel mechanism of RecA inhibition by the DinD protein. Fig. 4A illustrates the effect of the DinD protein (0.25 to 4 M) on the RecA-mediated DNA strand exchange when DinD is preincubated with the 20 M circular ssDNA substrate prior to the addition of a stoichiometric (1 RecA monomer per 3 nucleotides) concentration of RecA protein (to 6.7 M). The amount of nicked circular dsDNA product formed in 60 min was reduced in a DinD concentration-dependent manner and 2 M DinD was sufficient to completely inhibit the reaction, although significant inhibition of strand exchange products is observed by as little as 250 nM DinD protein (Fig. 4A). The amount of product over time between 60 and 120 min remained constant in the presence of DinD and the kinetics of early product formation were not affected by a suboptimal (0.5 M) DinD concentration (data not shown). Taken together, these data suggest that DinD may be competing with RecA in some manner. Because DinD protein can bind independently to the circular ssDNA substrate used in DNA strand exchange (Fig. 1A), it is possible that the inhibition of DNA strand exchange observed in Fig. 4A simply reflects a competition for DNA binding sites between the DinD protein and RecA filaments.
Based on the above observation that DinD does not readily compete with preformed RecA filaments for ssDNA (Fig. 2B), we hypothesized that DinD would exhibit minimal effects on the RecA-mediated DNA strand exchange if added to the reac-tion after RecA filaments have formed, i.e. Phase I of the reaction is complete (see Fig. 3). As Fig. 4B illustrates, the DinD protein is indeed able to hinder the amount of DNA exchange products catalyzed by RecA protein when added to preformed RecA filaments. Albeit, slightly higher concentrations of DinD protein are necessary to inhibit the reaction under these condi-  tions compared with when DinD protein is preincubated with the ssDNA (Fig. 4A). Assuming that RecA and DinD are simply competing for ssDNA binding sites, it would make sense that more DinD is required to displace RecA filaments already formed. However, DinD does not appear to affect RecA-ssDNA nucleoprotein filaments under similar concentration ratios of DinD to RecA, as measured by the hydrolysis of ATP (Fig. 2B). Therefore, the inhibition of RecA-mediated DNA strand exchange observed when DinD is added to preformed RecA filaments is likely due to the ability of DinD to bind to the duplex DNA substrate (Fig. 1B). This mechanism is explored further below.
The DinD Protein Specifically Acts to Halt RecA-mediated DNA Strand Exchange Intermediates-The fact that the DinD protein binds to duplex DNA evokes a mechanism by which a nonspecific duplex DNA-binding protein can inhibit DNA strand exchange merely by binding and sequestering the homologous DNA away from the RecA filament, preventing the DNA pairing reaction. This type of mechanism for DNA strand exchange regulation may not have a specific role in vivo and, thus would be unlikely to explain the need for DinD protein during the DNA damage response of E. coli. However, we were motivated to consider further the effect of DinD protein on RecA-promoted activities based on the data shown in Fig. 4C, which demonstrates that DinD is inhibitory to the completion of RecA-promoted DNA strand exchange when added 7 min after the initiation of homologous pairing, early in the second phase of the reaction (see Fig. 3). At this point in the reaction, RecA protein has paired a portion of the dsDNA molecules into plectonemic joints (40), as judged by the reaction intermediates that are stable to the deproteinization procedure used to stop the reaction (Fig. 4C, lane 2). The progression toward products mediated by RecA (for 60 min prior to deproteinization) observed in the absence of DinD (Fig. 4C, lane 3) is arrested in a DinD concentration-dependent manner when the DinD protein is added to this early stage of the reaction (Fig. 4C, lanes  4 -8). Higher concentrations of DinD protein (above 2 M) are needed to halt the progression of an on-going DNA strand exchange than when DinD is added before the initiation of the reaction (Fig. 4, A and B). However, these DinD concentrations are still lower than the 6.7 M RecA included in the reaction. It is possible that the DinD protein is simply binding to all duplex DNA still available after the initiation of DNA pairing but, we believe that a simple sequestration mechanism may not explain this result. In Fig. 5, we demonstrate the effect of sequestration, using the purified E. coli YebG protein. As described above, we have observed that YebG interacts with the duplex DNA (Fig.  1C), but not the ssDNA (data not shown) strand exchange substrate. The addition of YebG protein to the RecA-mediated DNA strand exchange reaction prior to the addition of the homologous duplex DNA (after RecA filaments have formed) impedes the formation of products (Fig. 5A). This inhibition of nicked, circular dsDNA strand exchange products is dependent on the YebG protein concentration and in fact, looks nearly identical to the same reaction in the presence of the DinD protein (Fig. 4B). However, when up to 4 M YebG protein is added to the RecA-mediated reaction 7 min after the initiation of DNA strand exchange, no inhibition is observed (Fig. 5B). It is possible that the activity of a nonspecific duplex DNA-binding protein, such as either YebG or DinD, can indeed prevent RecA-dependent homologous pairing by simply binding up the duplex DNA substrate and sequestering it away from the RecA filament (explored further below). However, once the DNA pairing reaction has begun, the duplex DNA associated with RecA filaments may not be readily accessible to other proteins. In fact, during DNA strand exchange, the duplex DNA is relatively protected from nuclease digestion under certain conditions (41)(42)(43). Nevertheless, Fig. 4C suggests that the DinD protein inhibits the progression of RecA-mediated DNA strand exchange intermediates after DNA pairing has initiated in addition to sequestering the duplex DNA substrate.
RecA Filament Activity Is Disrupted during DNA Strand Exchange by Addition of the DinD Protein-DinD protein appears to impede the progress of the RecA-promoted DNA strand exchange reaction when added to the reaction following the initiation of DNA pairing (Fig. 4C). We have established that the DinD protein does not affect RecA-ssDNA nucleoprotein filaments (Fig. 2B). We next explored the status of RecA filaments during DNA strand exchange in the presence of DinD by monitoring the ATP hydrolytic activity of the RecA protein throughout the reaction (Fig. 6A). For this reaction, the concentrations of RecA and ssDNA are lowered (to 2 and 4 M, respectively) relative to the DNA strand exchange reactions assessed by agarose gel electrophoresis (Fig. 4). Schutte and Cox (37) have established that when homologous duplex DNA is added to RecA-ssDNA filaments, there is a conformational change in the RecA protein that results in an ϳ30% drop in the rate of ATP hydrolysis, as seen in Fig. 6A, reaction 2. This immediate reduction in rate is indicative of paranemic DNA pairing by the RecA protein and is entirely dependent on the homology of the duplex DNA substrate. This is illustrated by adding duplex DNA to the reaction that is heterologous to the ssDNA molecule bound by RecA (Fig. 6A, reaction 3) (37). In this case, the ATP hydrolysis rate remains the same as when no duplex DNA is added to the RecA-ssDNA nucleoprotein filaments (Fig. 6A,  reaction 1). This indicates that RecA filaments do not interact with heterologous DNA in a manner that can bring about the DNA pairing-dependent conformational change in the RecA protein that results in a lower rate of ATP hydrolysis.
Throughout the course of a DNA strand exchange reaction as well as after the completion of the reaction, the lower rate of ATP hydrolysis catalyzed by the RecA protein described above was observed. RecA remains bound to the heteroduplex product of the DNA strand exchange reaction and indeed, the rate of ATP hydrolysis during strand exchange is comparable with the rate of ATP hydrolyzed by RecA protein bound to duplex DNA (38). Although the physical basis for this drop in ATP hydrolysis likely involves a RecA conformational change that is not completely understood (see "Discussion"), monitoring the changes in ATP hydrolysis during DNA strand exchange is used here as a tool to assay for changes in the filament status of RecA during the reaction as a function of added DinD protein. In Fig. 6A,  reaction 4 illustrates the effect of DinD protein when added to RecA filaments before the initiation of DNA strand exchange.
The steady-state rate of ATP hydrolysis is not reduced upon the addition of homologous duplex DNA (compare with reaction 2) to RecA-ssDNA filaments in the presence of 1 M DinD protein. The same result was obtained when up to 3 M DinD protein was included (data not shown). This is consistent with the duplex DNA binding activity (Fig. 1B) of the DinD protein sequestering the duplex DNA molecule away from the RecA filament when present prior to the addition of the duplex molecule, similar to the addition of heterologous duplex DNA in the absence of DinD (Fig. 6A, reaction 3). Consequently, paranemic or plectonemic DNA pairing does not occur and the typically lower rate of ATP hydrolysis by the RecA protein in the presence of homologous duplex DNA is not observed when DinD can access the duplex DNA prior to initiation of homologous DNA pairing (Fig. 6A, reaction 4).
Notably, when the DinD protein (1 M) is added to the RecApromoted DNA strand exchange after the reaction is initiated (and DNA pairing is begun, see inset to Fig. 6A), the rate of ATP hydrolysis exhibits a slow decline (Fig. 6A, reaction 5). This result is qualitatively similar to the net disassembly of RecA filaments observed when the RecX protein is added to RecA-ssDNA filaments (10). As described above, the ATP hydrolysis activity of RecA protein is DNA-dependent. Changes in the rate of ATP hydrolysis have been correlated to a change in the amount of RecA protein bound to DNA using direct methods such as electron microscopy, especially when these changes are brought on by another protein (10,11,44,45). Therefore, the time-dependent decline of ATP hydrolysis observed when the DinD protein is added to an ongoing DNA strand exchange reaction (Fig. 6A, reaction 5) likely reflects the disassembly of RecA from DNA. This disassembly hypothesis is explored further below. The immediate reduction in the rate of hydrolysis measured upon the addition of DinD suggests that at least the initial inhibition of RecA protein occurs rapidly. This explains why we observe no progression of DNA strand exchange intermediates by agarose gel electrophoresis once the DinD protein is added to an ongoing RecA-catalyzed reaction (inset to Fig.  6A).
The inhibition of RecA filaments (as measured by the decline in ATP hydrolysis) during the DNA strand exchange reaction suggests that the DinD protein may act specifically on the RecA protein rather than simply a competition for DNA binding sites. We tested for a competitive effect of the duplex DNA binding YebG protein on RecA filament dynamics during the RecA-promoted strand exchange reaction (Fig. 6B). The YebG protein is capable of the sequestration of duplex DNA when added to the DNA strand exchange reaction prior to the homologous duplex DNA (Fig. 6B, reaction 3), as described above for the DinD protein. However, if 2 M YebG protein is added to the DNA strand exchange reaction after the homologous duplex DNA, the typical 30% lower rate of hydrolysis catalyzed by RecA remains unchanged (Fig. 6B, reaction 4). The data illustrated by Fig. 6B strongly suggest that the duplex DNA binding activity of the YebG protein can sequester duplex DNA away from the RecA filament, presumably a nonspecific activity. Yet, the YebG protein cannot effectively compete with RecA protein filaments during an ongoing DNA strand exchange reaction. We considered the possibility that the YebG protein simply has a lower affinity for duplex DNA than DinD. However, we observed no negative effects of the YebG protein when added to an initiated RecA-mediated DNA strand exchange reaction up to a concentration 8-fold higher than the lowest concentration of DinD that is sufficient to observe RecA filament disassembly under the same conditions (data not shown). Based on the DinD and YebG duplex DNA binding assays (Fig.  1), we estimate the affinity of DinD for duplex DNA to be no more than 2-fold higher than that of YebG. It is therefore likely that the DinD-imposed inhibition observed in Fig. 4C is due to a specific effect on the RecA protein.
The RecA Protein Disassembled from Duplex DNA in the Presence of DinD Protein Is Competent to Bind ssDNA-We hypothesize that the reduction in the rate of RecA-catalyzed ATP hydrolysis observed when the DinD protein is added to an ongoing DNA strand exchange reaction (Fig. 6A, reaction 5) reflects the disassembly of RecA from DNA. Alternatively, the DinD protein could be mediating a conformational change in RecA that results in the reduction of ATP hydrolysis, without RecA disassembling from the DNA. Because RecA can compete with DinD for binding to a ssDNA substrate (Fig. 2), it is likely that much of the inhibition observed during strand exchange reflects a specific inhibition of RecA filaments interacting with more than one DNA strand (see "Discussion"). If DinD disassembles RecA from the strand exchange intermediate, RecA monomers should, in theory, then be available to bind any ssDNA substrate or product in solution. This is not observed during DNA strand exchange reactions in Fig. 6A, because the SSB protein present binds to the free DNA (46). Therefore, we assessed whether RecA protein inhibited by DinD was competent to bind ssDNA by omitting SSB (Fig. 7). Because the SSB protein has a postsynaptic role in DNA strand exchange (46), the RecA protein was loaded directly onto duplex DNA (Fig. 7, reaction 1) and challenged with DinD protein. RecA binding to duplex DNA is slow under standard reaction conditions, thus the pH of the solution was lowered to 6 (from 7.5) to facilitate nucleation (47). For this reaction, the RecA protein (1.2 M) was stoichiometric with the duplex DNA (X174 nicked circular dsDNA) binding sites (3.6 M base pairs) to essentially eliminate any excess RecA protein in solution. In this case, the rate of ATP hydrolysis abruptly declines upon the addition of DinD protein (Fig. 7, reactions 2, 5, and 6). When poly(dT) ssDNA was subsequently added to that same reaction, the RecA recovers immediately, again hydrolyzing ATP (Fig. 7, reactions 5 and 6) at a rate dependent on the concentration of ssDNA added. When ssDNA is added to RecA-dsDNA filaments in the absence of DinD, only a slight increase in hydrolysis was observed (Fig. 7, reactions 3 and 4). Taken together, these data indicate that the RecA is displaced from duplex DNA in the presence of DinD and is free to bind the available ssDNA.
RecA Filaments Disassemble from the Product of DNA Strand Exchange Upon the Addition of DinD Protein-RecA filaments formed on ssDNA appear resistant to the negative effects of DinD observed during DNA strand exchange. Furthermore, RecA filaments formed directly on duplex DNA are very sensitive to DinD (Fig. 7). Therefore, we hypothesize that DinD functions to disrupt a different state of RecA filaments than that formed on ssDNA. The RecA filaments appear to be more dynamic during the process of DNA strand exchange, with protomers moving into and out of the filament (36), perhaps at the branch points (48,49). DinD protein could, in theory be acting on the RecA protein during the branch migration phase of DNA strand exchange, where RecA filaments are likely in contact with three DNA strands. A simpler explanation is that the DinD protein acts on filaments when the RecA protein exists in a "P" conformation, a functional state of RecA filaments bound to duplex DNA (see "Discussion"). To test this possibility, we added the DinD protein to the RecA-promoted DNA strand exchange reaction at different times during the progression of the reaction, 7, 15, and 30 min after initiation and after the reaction was completed, 45 min after initiation (Fig. 8). The reaction progress is illustrated by the agarose gel shown in the inset to Fig. 8A. At 7, 15, and 30 min, DNA strand exchange is still in progress, as judged by the substrate molecules remaining after deproteinization. By 45 min after the DNA strand exchange reaction is initiated, all linear duplex DNA substrate have been converted to nicked circular dsDNA product. The RecA protein remains bound to the duplex DNA product of the strand exchange reaction (38,39). Therefore, by adding the DinD protein to the strand exchange reaction at 45 min we were able to monitor the effect of DinD protein on RecA filaments bound to duplex DNA at physiological pH. Fig. 8 illustrates that the DinD protein efficiently disassembles RecA protein filaments that are bound to the duplex DNA product of DNA strand exchange as well as RecA filaments actively progressing DNA strand exchange intermediates. We carefully titrated the DinD protein (0.25-2 M) into these DNA strand exchange experiments, adding the DinD protein 7, 15, 30, and 45 min after strand exchange was initiated. The results of two DinD concentrations tested are shown in Fig. 8 (0.25 M, panel  A; 1 M, panel B). We reasoned that a protein titration would suggest whether the DinD has a greater impact on RecA filaments during the progression of DNA strand exchange (Fig. 8,  7, 15, and 30 min) or on RecA filaments bound to the duplex DNA product of strand exchange (Fig. 8, 45 min). At a given DinD protein concentration, the amount of RecA protein that disassembles from DNA is approximately the same, whether DinD was added during the progression of the reaction (7, 15, or 30 min) or after the completion of the reaction (45 min). We note that the rate of RecA disassembly is consistently slower, albeit slightly, when the DinD protein is added to the early stage of this reaction (7 min) at concentrations above 0.5 M (see Fig. 8B) than when added later in the reaction. We are unable to explain this difference using the current assay. Nevertheless, there is minimal difference in the inhibition of RecA filaments observed when the DinD protein is added to the last two distinct phases (see Fig. 3) of the DNA strand exchange reaction, the intermediate stage when RecA protein is in contact with three DNA strands and the product stage when RecA is bound to duplex DNA. We conclude that DinD is a potent inhibitor of RecA protein bound to two or more strands of DNA.

DISCUSSION
The primary conclusion of this study is that the E. coli DinD protein is a novel inhibitor of the RecA-mediated DNA strand exchange reaction. The target of this inhibition is not the RecA-ssDNA nucleoprotein filament, rather the DinD protein specifically affects RecA filaments bound to two or more DNA strands. Consequently, RecA filaments actively promoting DNA recombination or RecA filaments bound to duplex DNA are disassembled in the presence of the DinD protein (Figs. 6 -8) resulting in the desistance of branch migration (Figs. 4C and 6A, inset). To date, no other protein involved in RecA regulation has been shown to efficiently disassemble RecA filaments bound to multiple DNA strands whereas, at the same time, exhibiting no clear effect on RecA-ssDNA filaments. Therefore, we propose that the negative effects of the DinD protein are specific to the conformational state of the RecA protein that predominates when RecA is bound to duplex DNA.
A functional P state, one of four distinct functional states of the RecA protein has been described in detail by Cox and coworkers (50 -52). This P state of RecA occurs when RecA pro- tein is bound to two strands of DNA and is structurally distinct from the RecA bound to ssDNA. The characteristics of the P state of the RecA protein includes a lower rate of ATP hydrolysis, a higher degree of cooperativity between monomers, and increased filament dynamics (50). The P state is proposed to predominate for RecA protein that is interacting with two strands of DNA, whether it binds directly to duplex or as a result of homologous pairing. The transition between the conformational state of RecA bound to a ssDNA molecule and the P state occurs immediately upon the addition of duplex DNA that is homologous in sequence to the ssDNA bound by RecA (Fig. 9), as observed by the immediate reduction in the rate ATP hydrolysis catalyzed by the RecA protein (Fig. 6, reaction 2). The magnitude of the drop in ATP hydrolysis is dependent on the amount of homology present in the duplex DNA, and likely occurs faster than the appearance of plectonemic DNA strand exchange intermediates (37). Therefore, it is possible that RecA protein adopts the P conformational state along the region of the ssDNA-nucleoprotein filament that interacts with the duplex DNA prior to strand switching. These paranemic regions (see Fig. 9) may provide an explanation for the fact that the dsDNA substrate during strand exchange is not readily accessible to other proteins, such as nucleases, under some conditions (41)(42)(43) and why a nonspecific duplex DNA-binding protein (such as YebG, Fig. 6B, reaction 4) might not be able to halt branch migration through the binding of dsDNA not yet incorporated into the RecA filament, as illustrated in Fig. 9.
We have established that the DinD protein can inhibit the RecA-promoted DNA strand exchange reaction in vitro by three different mechanisms: 1) DinD can interfere with RecA nucleation ( Fig. 2A) through its binding to ssDNA (Fig. 1A), thus inhibiting phase 1 of the reaction (Fig. 3). 2) DinD can interfere with the pairing of homologous duplex DNA by binding (Fig. 1B) and sequestering (Fig. 6A) the dsDNA molecule away from RecA-ssDNA filaments, thus inhibiting phase 2 of the reaction. Importantly, DinD cannot inhibit phase 2 of this reaction through the disassembly of RecA-ssDNA filaments (Figs. 2B and 6, reaction 4), making DinD distinctly different from RecX, UvrD, and PcrA proteins (8 -10). 3) DinD can halt the progression of an initiated DNA pairing reaction (Fig. 4C) by promoting the disassembly of the RecA protein (Figs. 6 and 7). We argue that the first and second mechanism listed here, interference, can be accomplished in vitro by the action of DNAbinding proteins. For example, SSB protein bound to ssDNA competitively interferes with RecA filament formation (5) and binding of the RgdC protein to duplex DNA effectively interferes with homologous DNA pairing by RecA (7). We have also shown this to be true of the YebG protein (Fig. 5A). However, the simple mechanism of interference does not appear sufficient to explain the disassembly of RecA protein during an ongoing DNA strand exchange reaction imposed by the DinD protein. The duplex DNA binding ability of YebG protein (Fig.  1C) can effectively inhibit RecA-mediated DNA strand exchange when present prior to the addition of the homologous duplex substrate (Figs. 5A and 6B, reaction 3). Conversely, YebG protein does not inhibit and has no measurable impact on the stability of RecA filaments if added to a RecA-promoted reaction that has already initiated (Figs. 5B and 6B, reaction 4). As noted above, RdgC can also sequester duplex DNA away from a RecA-ssDNA filament to inhibit strand exchange (7) and the RuvA and RuvB DNA translocases can disassemble RecA filaments from supercoiled dsDNA (53). However, RecA filament stability during DNA strand exchange was not measured in either of those studies. The observation that the DinD protein is able to disassemble the RecA protein during an ongoing DNA strand exchange reaction has not specifically been reported for other proteins with a described or predicted role in regulating RecA function.
The disassembly of RecA filaments bound to more than two strands of DNA, observed here by the DinD protein, could have multiple roles in bacterial cells. RecA filaments are likely bound to duplex DNA at various times in vivo. Even though RecA nucleation onto duplex DNA is slow at physiological pH, direct duplex DNA binding is intrinsic to the RecA protein (47,54) and it has been suggested that RecA mutants that bind better to duplex DNA than the wild-type protein may have imbalanced levels of SOS (55). Also, RecA protein that has nucleated onto a ssDNA gap can also extend into the duplex region beyond the gap (56). And finally, RecA protein bound to the heterodu- RecA-ss-DNA nucleoprotein filaments interact with a homologous duplex DNA molecule, pairing the complementary strand (gray) and displacing the identical strand (black). The RecA protein adopts a P conformational state (rectangles) that is distinct from the state when bound only to ssDNA or free RecA monomers (both depicted as ovals). The plectonemic joint molecule intermediate of this reaction contains a region of heteroduplex behind the branch and a region of paranemic association between RecA-ssDNA filaments and the substrate duplex molecule ahead of the branch. Branch migration (to the right) would continue through to the end of the duplex. For assays in vitro, the products are the displaced linear single strand and the heteroduplex molecule (see Fig. 3). DinD protein can affect both the presynaptic and postsynaptic stages of this RecA-mediated process. DinD can prevent synapsis through the sequestration of the duplex DNA substrate. This sequestration effect can also be imposed by a nonspecific duplex DNA-binding protein. Notably, however, DinD can also mediate the disassembly of postsynaptic RecA filaments, whether bound to the heteroduplex product or involved in paranemic associations, perhaps by targeting the P state of RecA filaments. plex DNA produced by homologous DNA recombination (38,39) is likely, regardless of the pathway in which RecA is employed. In each of these circumstances, removing RecA from duplex DNA would aid in the recycling of the recombinase protein. Importantly, this would also ensure that RecA would not interfere with events downstream of homologous DNA pairing, such as the extension of the 3Ј OH for recombination-dependent synthesis, branch migration, and/or Holliday junction resolution.
In addition to the disassembly of the RecA protein bound to duplex DNA as outlined above, there may be a role for halting the branch migration of ssDNA strand invasion products. This could be accomplished by disassembling RecA from DNA strand exchange intermediates, as we have described here for the DinD protein. It has been suggested that RecA-mediated plectonemic joints may be barriers to chromosomal replication (57). In eukaryotic systems, mediators of the RecA orthologs, Rad51 and DMC1, have a role in determining whether recombination leads to crossover or non-crossover events. Crossover events can result from typical DNA double-strand break repair pathways that include extensive branch migration, followed by double Holliday junction resolution. Noncrossover events can result from pathways such as synthesisdependent DNA strand annealing, where Rad51 pairing intermediates would need to be disassembled. Recently, it has been suggested that the meiosis-specific DMC1 protein is more resistant to disassembly at a D-loop by regulator proteins (58), thereby continuing through the DNA doublestrand break repair pathway that promotes genetic exchange between homologous chromosomes. Conversely, Rad51 is more sensitive to regulation (58), thereby easily disassembled for the synthesis-dependent DNA strand annealing pathway that does not lead to crossovers. Future work will be necessary to determine whether the DinD protein has a role in affecting RecA filaments to regulate the recombinational DNA repair pathway choice in bacterial systems.